Finch / FinWorkBench

Composite financial workflow · artifact evaluationReportSource task tags: Validation / Review · Calculation · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Complete the validation and indicator calculations as follows: on the Balance Sheet, add a control to ensure TOTAL ASSETS equals TOTAL LIABILITIES AND EQUITY; on the Income Statement (Revenue & Expenses), add an Equity Roll Forward Test to reconcile equity movement and highlight any differences.

{
  "files": [
    "0_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Calculation · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Compute the “sum of A” and “sum of B” in the last rows of the Income sheet for each Financial Indicator Component, using data from all sheets in the workbook.

{
  "files": [
    "1_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Validation / Review · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Correct the year header on the Five Year Review tab (set the first year to 2000) and bring that tab into alignment with the other worksheets by updating any figures that are inconsistent.

{
  "files": [
    "2_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Cross-sheet/file Retrieval · Data Entry / Import · Calculation · Validation / Review

Original English instruction · agent OS and apps are not declared.

Complete the missing data in the 'Five Year Review' worksheet of the first given xlsx file, ensuring the five-year figures across fund balances, operating revenues, and operating expenditures are fully populated and consistent with the rest of the financial statements. Leave blank if no reference is available for specific parts.

{
  "files": [
    "3_ref_0.xlsx"
  ],
  "text": ""
}

{
  "review_flags": [
    "first_source_excel_only"
  ],
  "review_notes": [
    "The content builder selects only the first Excel file for each source/reference/output side. IDs 3 and 108 each have two source XLSX inputs; ID 108 explicitly requests use of the second. Excel preprocessing recalculates files but returns no textual evidence entries. Record source-visibility limits; verify all required inputs appear in any future judge payload. This does not establish task infeasibility or a measured grading error."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E01",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/build_prompt/content_builder/content_builder.py#L171-L261",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/build_prompt/content_builder/utils.py#L64-L75",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_excel.py#L305-L332"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Cross-sheet/file Retrieval · Data Entry / Import · Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared.

Using the images and the current workbook, please complete the missing entries on the ‘Five Year Review’ sheet. Populate the ‘FIVE YEARS IN REVIEW – FINANCIAL DATA’ lines so the five-year figures and totals are filled in and consistent with the rest of the file. Leave blank if no reference is available for specific parts.

{
  "files": [
    "4_ref_0.xlsx"
  ],
  "text": ""
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationInvestment: CreditSource task tags: Data Entry / Import · Structuring / Formatting

Original English instruction · agent OS and apps are not declared.

Transcribe the content from the pdf/image into the Excel file.

{
  "files": [
    "5_ref_0.xlsx"
  ],
  "text": ""
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Calculation · Summary / Visualization

Original English instruction · agent OS and apps are not declared.

Please write a structured economic analysis report based on the table data I provide. Requirements: Base the analysis only on the table content; do not introduce external knowledge. Identify trends, changes, turning points, and fluctuation patterns during crisis periods in the data. Analyze relationships between different indicators, such as the connection between time periods and specific indicator changes, or whether certain event years exhibit significant volatility. Integrate your findings into a formal economic report–style narrative and include clear visual charts, for example: “During the period of xxx, we observe that...” “The data indicate that the institution exhibits...” The report must include: A summary of trends Characteristics of special years Inferred economic behavior patterns The report should be no more than 300 words.

{
  "files": [
    "6_ref_0.xlsx"
  ],
  "text": "Figure 1.4 shows that, for low- and middle-income countries over 1971–2023, new World Bank lending (IBRD and IDA, including IDA grants) tends to rise as a share of GNI when GNI growth declines. During major downturns – such as the 2008–09 financial crisis and the COVID-19 pandemic – the ratio of IBRD and IDA lending (and IDA grants) to GNI increases noticeably even as economic growth turns sharply negative. This pattern, which recurs around several episodes of weak growth and external shocks, indicates that World Bank lending has played a countercyclical and stabilizing role, expanding in periods of stress to help cushion dramatic drops in economic activity in eligible low- and middle-income economies."
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Summary / Visualization · Translation

Original English instruction · agent OS and apps are not declared.

The report should be translated into English, maintaining a neat layout, and a docx version of the report should be generated.

{
  "files": [
    "7_ref_0.docx"
  ],
  "text": ""
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReport · Predictive ModelingSource task tags: Web Search · Data Entry / Import · Structuring / Formatting · Calculation · Validation / Review

Original English instruction · agent OS and apps are not declared.

I am studying how partisan standoffs influence administrative efficiency and the macro-economy, so I need a quantitative overview of every federal government shutdown.\n\nPlease output an xlsx file with Sheet1 as the RawData sheet, with the columns in this order:\nStart Date, End Date, Duration (days), President, Speaker of the House, Senate Majority Leader, Furloughed Employees, Estimated Loss (USD million) , Main Disputed Provisions\n\nRequirements:\n1. Cover every officially recorded federal government shutdown between October 1976 and December 2024 (including October 1976 and December 2024).\n2. Date should be formatted as YYYY-MM-DD.\n3. Record the President, House and Senate leaders during the shutdown period.\n4. For furloughed employees and estimated loss, give the exact number. If there is no exact number or data is unavailable, fill in with \"–\".\n\nDon't ask me any questions, just output the results according to the columns without omitting cells arbitrarily.\n\nBased on the RawData sheet, complete the following 4 subtasks:\n\nSubtask 1 - Sheet2_Presidential_Era: Aggregate shutdowns by President to analyze patterns across different administrations.\nColumns: President, Number of Shutdowns, Total Days, Average Duration (days), Longest Shutdown (days), Total Furloughed Employees, Total Estimated Loss (USD million), Era Severity Rating\nAverage Duration with 1 decimal. Era Severity Rating: \"High Frequency & High Impact\" if ≥5 shutdowns AND ≥30 total days; \"High Frequency\" if ≥5 shutdowns only; \"High Impact\" if ≥30 total days only; \"Moderate\" if ≥2 shutdowns; \"Low\" otherwise. For totals: \"N/A\" if all source values are \"–\", otherwise sum available numeric values. Sort by presidential chronological order (by order of first appearance in RawData).\n\nSubtask 2 - Sheet3_Timeline_Context: Add chronological context and temporal analysis.\nColumns: Start Date, End Date, Duration (days), Year, Decade, President, Days Since Previous Shutdown, Cumulative Shutdown Count, Duration Category\nYear: extract from Start Date (integer). Decade: \"1970s\", \"1980s\", \"1990s\", \"2000s\", \"2010s\", \"2020s\". Days Since Previous Shutdown: days between previous shutdown's End Date and current Start Date (integer), \"First Event\" for row 1. Cumulative Shutdown Count: sequential number 1, 2, 3... (integer). Duration Category: \"Extended (≥14 days)\" if ≥14, \"Moderate (5-13 days)\" if 5-13, \"Brief (<5 days)\" if <5.\n\nSubtask 3 - Sheet4_Data_Validation: Assess data completeness for each shutdown event.\nColumns: Start Date, President, Has Furloughed Data, Has Loss Data, Has Provisions Data, Data Completeness (%), Quality Rating, Missing Fields\nHas [Field] Data: \"Yes\" if data exists, \"No\" if \"–\". Data Completeness (%): (count of \"Yes\" fields / 3) × 100 (1 decimal). Quality Rating: \"Complete\" if 100%, \"Good\" if ≥66.7%, \"Partial\" if ≥33.3%, \"Poor\" if <33.3%. Missing Fields: comma-separated list like \"Furloughed, Loss\" or \"None\".\n\nSubtask 4 - Sheet5_Impact_Classification: Classify shutdown severity based on multiple impact dimensions.\nColumns: Start Date, Duration (days), President, Furloughed Employees, Estimated Loss (USD million), Duration Impact, Economic Impact, Workforce Impact, Overall Severity Score, Overall Impact Classification\nDuration Impact: \"Severe\" if ≥14 days, \"Moderate\" if 5-13, \"Minor\" if <5. Economic Impact: \"Unknown\" if N/A or \"–\", \"Major\" if ≥1000M, \"Significant\" if ≥100M, \"Moderate\" if ≥10M, \"Minor\" if <10M. Workforce Impact: \"Unknown\" if N/A or \"–\", \"Massive\" if ≥500K, \"Major\" if ≥100K, \"Moderate\" if ≥10K, \"Limited\" if <10K. Overall Severity Score: Duration component (max 40 points) + Economic component (max 30 points) + Workforce component (max 30 points), result with 1 decimal. Overall Impact Classification: \"Critical\" if ≥70, \"High\" if ≥50, \"Moderate\" if ≥30, \"Low\" if ≥10, \"Minimal\" if <10.

{
  "files": [
    "8_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReport · Predictive ModelingSource task tags: Web Search · Data Entry / Import · Calculation · Structuring / Formatting

Original English instruction · agent OS and apps are not declared.

For an infrastructure-finance paper, I need to benchmark capital intensity of large offshore wind assets. List every offshore wind farm in European waters that was fully commissioned (i.e., all turbines installed and formally entered commercial operation) from 2010-01-01 to 2024-12-31 and whose nameplate capacity is >= 300 MW. Ignore projects still under construction or phases that are only partially energized / not yet formally fully commissioned as of December 31, 2024.\n\n[Data Source and Methodology Notes]\n1. Data Source: Please use 4C Offshore or WindEurope database records as of October 31, 2024.\n2. Commissioning Boundary (Strict Definition): Projects must have formally announced full commissioning before December 31, 2024. Projects that only completed turbine installation but have not yet formally achieved full commissioning (e.g., Baltic Eagle completed installation in 2024 but full commissioning in July 2025) should NOT be included.\n3. Capacity Convention:\n - Hollandse Kust Zuid 1-2: 770 MW / 70 turbines\n - Hollandse Kust Zuid 3-4: 759 MW / 69 turbines (official KEC convention)\n - Borkum Riffgrund 2: 450 MW (developer convention)\n - Seagreen: 1075 MW (developer official convention)\n4. Combined Projects: Combined projects (e.g., Gode Wind 1+2, Thornton Bank C-Power phases) are treated as single wind farms.\n\nPlease output an xlsx file with Sheet1 as the RawData sheet, with the columns in this exact order:\nWind Farm, Sea / Basin, Capacity (MW), Turbines Number, Turbine Model, Commissioning Year, Owner / Operator\nFill missing fields with \"NA\".\nDon't ask me any questions, just output the results according to the column without omitting cells arbitrarily.\n\nBased on the RawData sheet, complete the following 4 subtasks:\n\nSubtask 1 - Sheet2_Turbine_Unit_Economics: Analyze turbine-level economics and efficiency metrics.\nColumns: Wind Farm, Capacity (MW), Turbines Number, Average Turbine Capacity (MW), Capacity Density (turbines/100MW), Scale Category, Turbine Size Category\nAverage Turbine Capacity (MW) = Capacity (MW) / Turbines Number (2 decimals). Capacity Density (turbines/100MW) = (Turbines Number / Capacity (MW)) x 100 (2 decimals). Scale Category: \"Gigawatt-Scale\" if Capacity >= 1000 MW, \"Large-Scale\" if >= 600, \"Mid-Scale\" if >= 400, \"Standard-Scale\" otherwise. Turbine Size Category: \"Ultra-Large (>=10MW)\" if Average >= 10, \"Large (7-10MW)\" if >= 7, \"Medium (5-7MW)\" if >= 5, \"Standard (<5MW)\" otherwise.\n\nSubtask 2 - Sheet3_Geographic_Market: Aggregate projects by sea/basin to analyze regional market structure and concentration.\nColumns: Sea / Basin, Number of Projects, Total Capacity (MW), Average Project Size (MW), Largest Project (MW), Market Share (%), Market Development\nNumber of Projects: COUNT of projects per region (integer). Total Capacity (MW): SUM of all capacities in region (2 decimals). Average Project Size (MW): Total / Number (2 decimals). Largest Project (MW): MAX capacity in region (2 decimals). Market Share (%): (Regional Total / Grand Total) x 100 (2 decimals). Market Development: \"Concentrated\" if <= 2 projects, \"Moderate\" if 3-5, \"Diversified\" if >= 6. Sort by Total Capacity (MW) descending.\n\nSubtask 3 - Sheet4_Technology_Manufacturer: Extract manufacturer from turbine model strings and classify technology generation.\nColumns: Wind Farm, Turbine Model, Commissioning Year, Manufacturer, Manufacturer Type, Unit Capacity (MW), Technology Generation\nManufacturer: Extract from Turbine Model string (e.g., \"Siemens\", \"Vestas\", \"GE\", \"MHI Vestas\", \"AREVA\", \"Senvion\", \"Adwen\", \"REpower\", \"Bard\", \"Other\"). Manufacturer Type: \"Major OEM\" if Siemens/Vestas/GE, \"Established Player\" if MHI Vestas/Senvion/AREVA, \"Specialized/Regional\" otherwise. Unit Capacity (MW): Primarily extract rated capacity from model name (e.g., \"SWT-6.0-154\" → 6.0 MW), even if it differs from farm capacity / turbine count; 2 decimals. Technology Generation: \"Next-Gen (>=10MW)\" if >= 10, \"Gen 4 (8-10MW)\" if >= 8, \"Gen 3 (6-8MW)\" if >= 6, \"Gen 2 (4-6MW)\" if >= 4, \"Gen 1 (<4MW)\" otherwise.\n\nSubtask 4 - Sheet5_Operator_Portfolio: Identify major operators by parsing ownership strings and calculate portfolio metrics. Show top 15 operators only.\nColumns: Operator Name, Number of Projects, Total Capacity (MW), Average Project Size (MW), Portfolio Rank, Market Share (%)\nParse Owner/Operator field (split by \"/\" separator) to identify individual operators. Normalize common operator names (e.g., \"Orsted\"/\"orsted\" -> \"Ørsted\"). Number of Projects: Count of distinct projects in which the operator participates (integer, not split for joint ownership). Total Capacity (MW): SUM of capacities, split equally among co-owners (2 decimals). Average Project Size (MW): Total / Number (2 decimals). Portfolio Rank: 1 = highest total capacity (integer). Market Share (%): (Operator Total / Sum of Top 15) x 100 (2 decimals). Sort by Total Capacity descending, include only top 15 operators.

{
  "files": [
    "9_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Calculation · Financial Modeling

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Per the headers and established formula logic, populate formulas for columns X through AH so the timing model’s performance statistics for 2013–2025 are complete and consistent with the existing approach.

{
  "files": [
    "10_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Summary / Visualization

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Add a NAV chart above the Strategy Performance table, plotting two series: Rotation NAV and Benchmark. Use a red line with red markers for Rotation NAV and a gray line with gray markers for the Benchmark; start both series at a base value of 1, and set the time axis to show annual ticks.

{
  "files": [
    "11_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Structuring / Formatting · Validation / Review · Summary / Visualization · Calculation · Financial Modeling

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Per the red parameters and the Method 1/Method 2 guidance noted in H8 and H9, complete the formulas in columns T and U (starting from1), Note that the starting point for the formulas in columns T and U is 1, representing the initial signal to hold Index 1. In the formulas for columns T and U, 1 represents the signal to hold Index 1, -1 represents the signal to hold Index 2, and 0 represents the signal to make no change. Then complete column I. The method selection in B6 should drive the model so that all cells and charts refresh consistently when switching between methods.

{
  "files": [
    "12_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Complete the missing formulas in columns E and F on the sheet Holding Period Return Analysis. This task uses the yield curve on the given date to compute the annualized return over the specified holding period for a pair of virtual bonds. The calculation steps should follow the logic shown in column D, and the yellow-highlighted cells are input parameters. The model should update the virtual bonds’ holding-period return dynamically when these parameters change. Compute the “Initial Full Price” by setting the coupon rate equal to the current yield. Determine the “Remaining Maturity at the End of the Holding Period” based on the parameters in the table. Use the cash-flow discounting formula and the corresponding parameters to compute the “Final Full Price.”

{
  "files": [
    "13_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Financial Modeling · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Suppose we need to hold a 0.5-year AA(2) municipal investment bond. Using the model, compare the holding-period returns of the 2-year and 4-year tenors. It is known that over the next half year, the 1-year yield will rise by 2 bps, and the 5-year term spread relative to the 1-year will widen by 3 bps.

{
  "files": [
    "14_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Structuring / Formatting · Translation · Summary / Visualization

Original English instruction · agent OS and apps are not declared.

Translate all Chinese text in this Excel workbook (including sheet names, headers, cell contents, formulas, charts, etc) into English and save the translated version as a new workbook.

{
  "files": [
    "15_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Calculation · Summary / Visualization

Original English instruction · agent OS and apps are not declared.

Continue writing the report based on the provided spreadsheet data, and generate the subsequent part of the report, saving it as a PDF file that maintains the same format as the original report. Requirements: Use the data from each spreadsheet provided to supplement the report. Update the report section by section based on the spreadsheet data.

{
  "files": [
    "16_ref_0.pdf"
  ],
  "text": ""
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationAccounts Payable and ReceivableSource task tags: Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared.

For the 10th of the year 2023, what are the Fee IDs applicable to Belles_cookbook_store? Answer must be a list of values in comma separated list, eg: A, B, C. If the answer is an empty list, reply with an empty string. If a question does not have a relevant or applicable answer for the task, please respond with 'Not Applicable'

{
  "files": [],
  "text": "741, 709, 454, 813, 381, 536, 473, 572, 477, 286"
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge",
    "source_csv_json_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input.",
    "The preprocessing chain handles PDF, Markdown, Word, Excel and images, with no CSV/JSON handler. The content builder attaches source content through the first Excel path; its general-side builder covers reference/output, not source files. IDs 17, 18 and 73 contain CSV/JSON inputs. Do not claim the current default judge directly inspects those structured inputs. Keep reference text and original source links; assess coverage before any benchmark run."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    },
    {
      "issue_id": "FINCH-E03",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L113-L126",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/build_prompt/content_builder/content_builder.py#L205-L261"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationAccounts Payable and ReceivableSource task tags: Cross-sheet/file Retrieval · Calculation

Original English instruction · agent OS and apps are not declared.

In January 2023 what delta would Belles_cookbook_store pay if the relative fee of the fee with ID=384 changed to 1? Answer must be just a number expressed in EUR rounded to 6 decimals. If a question does not have a relevant or applicable answer for the task, please respond with 'Not Applicable'

{
  "files": [],
  "text": "-0.940000"
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge",
    "source_csv_json_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input.",
    "The preprocessing chain handles PDF, Markdown, Word, Excel and images, with no CSV/JSON handler. The content builder attaches source content through the first Excel path; its general-side builder covers reference/output, not source files. IDs 17, 18 and 73 contain CSV/JSON inputs. Do not claim the current default judge directly inspects those structured inputs. Keep reference text and original source links; assess coverage before any benchmark run."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    },
    {
      "issue_id": "FINCH-E03",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L113-L126",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/build_prompt/content_builder/content_builder.py#L205-L261"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Web Search · Data Entry / Import · Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared.

Please organize the following data for Apple, Amazon, Google, Microsoft, and Netflix over the five-year period from 2020 to 2024: Total revenue (in billions of US dollars), Operating profit according to International Financial Reporting Standards (in billions of US dollars), Operating profit margin, and Free cash flow (FCF, in billions of US dollars). Please quote the data from the financial reports of the company. Mark the data that cannot be found with \"-\". The value should be specified to two decimal places.\n\nPlease output an xlsx file with Sheet1 as the RawData sheet, with the following column names in sequence:\nCompany, Year, Total revenue (Billion), Operating profit (Billion), Operating profit margin, FCF (Billion)\n\nDon't ask me any questions, just output the results according to the column without omitting cells arbitrarily.\n\nBased on the RawData sheet, complete the following 4 subtasks:\n\nSubtask 1 - Sheet2_YoY_Growth: Calculate year-over-year growth rates for each company.\nColumns: Company, Year, Total Revenue(Billion), Revenue YoY Growth(%), Operating Profit(Billion), Operating Profit YoY Growth(%), FCF(Billion), FCF YoY Growth(%)\nFirst year (2020) for each company should show N/A for growth rates. All values with 2 decimals.\n\nSubtask 2 - Sheet3_Ranking_By_Year: Reorganize data by year and rank companies within each year.\nColumns: Year, Company, Total Revenue(Billion), Revenue Rank, Operating Profit(Billion), Operating Profit Rank, FCF(Billion), FCF Rank, Market Leader\nWithin each year, sort companies by revenue (highest to lowest). Ranks are integers 1-5. Market Leader: \"Yes\" for the company with highest revenue in each year, \"No\" for others.\n\nSubtask 3 - Sheet4_Five_Year_Summary: Aggregate 5-year statistics for each company.\nColumns: Company, Total Revenue 5-Year Sum(Billion), Average Annual Revenue(Billion), Revenue CAGR 2020-2024(%), Total FCF 5-Year(Billion), Average Operating Margin(%), Operating Profit CAGR 2020-2024(%), Growth Category\nCAGR formula: ((Ending Value / Beginning Value)^(1/4) - 1) × 100 (4 years from 2020 to 2024). Growth Category: \"High Growth\" if Revenue CAGR ≥ 15%, \"Moderate Growth\" if ≥ 5%, \"Stable Growth\" otherwise. All percentages with 2 decimals.\n\nSubtask 4 - Sheet5_Financial_Health: Assess financial health based on 2024 data.\nColumns: Company, Revenue 2024(Billion), Operating Margin 2024(%), FCF 2024(Billion), Profitability Score, Cash Generation Score, Scale Score, Total Health Score, Health Rating, Risk Level\nProfitability Score: 40 if Operating Margin ≥ 30%, 30 if 20-30%, 20 if 10-20%, 10 if < 10%. Cash Generation Score: 30 if FCF ≥ 70B, 20 if 30-70B, 10 if < 30B. Scale Score: 30 if Revenue ≥ 500B, 25 if 350-500B, 20 if 200-350B, 10 if < 200B. Total Health Score = sum of above three scores. Health Rating: \"Excellent\" if Total ≥ 85, \"Good\" if 70-84, \"Fair\" if 55-69, \"Poor\" if < 55. Risk Level: \"Low Risk\" if Total ≥ 85, \"Medium Risk\" if 55-84, \"High Risk\" if < 55.

{
  "files": [
    "19_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPlanning and BudgetingSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

At the end of the table, add two summary rows labeled 'Total Disbursements' and 'Net Receipts (Disbursements)'.

{
  "files": [
    "20_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPlanning and BudgetingSource task tags: Data Entry / Import · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Add a weekday line directly below the date headers and update the 12/31/2001 (Mon) column. For that day, there are no “Receipts”; record disbursements of $1,980,800 to Calpine (Power Purchases) and $100,000 to an unspecified vendor (Gas Purchases). Under Enron Facility Services, enter $3,101,855 for “$2.5 per day” and -$2,081,386 for “estimate receipt”; in Personnel, EES is $584,500; leave all other items as “-”.

{
  "files": [
    "21_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Validation / Review · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Please review the pivot table on the Replacement Cost sheet against the LiquidationValue sheet to confirm they are consistent. If there are discrepancies, update the Replacement Cost pivot to reflect the LiquidationValue data.

{
  "files": [
    "22_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Calculation · Validation / Review

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Using the pivot table’s monthly Nominal Dollars as the base, compute a fixed 50% Margin Requirement – Fixed NYMEX for each month, then calculate the Cumulative Margin Required as 2x the current month’s margin plus the sum of margins from the next month through the final month. On this balance, apply a 25% annual rate converted to monthly (÷12) to derive the Monthly Cost of Capital – NPW, discount at 3% (APR) annually to obtain monthly present values, and sum those to report the Total Cost of Credit to NPW; if needed, you may validate Nominal Dollars using price × quantity, but for this run use Nominal Dollars.

{
  "files": [
    "23_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Structuring / Formatting · Financial Modeling

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Add a new worksheet named "Scenario3" to the same workbook, mirroring the structure, row/column layout, monthly detail table, and chart area of "Scenario1". For Scenario3, update the hedging assumptions to a balanced allocation: 10-Yr 25%, 5-Yr 20%, 1-Yr 15%, May-Sep 20%, Q3 15%. Keep the note "Maximum Monthly Average Short Position to Cover (July Peak) = 30,508 MW" unchanged; only the new sheet should be added, and formulas may be used within it.

{
  "files": [
    "24_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationProcurement and Sales · Trading and Risk ManagementSource task tags: Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared.

Using the daily data for August from the Eron (NC) Service Invoice, verify whether they reconcile with the monthly totals for Net Receipts – Allocated UA4 and Deliveries in the Imbalance Statement. Then compute the final month’s cumulative imbalance.

{
  "files": [],
  "text": "The monthly Net Receipt of 714,490 matches the sum of the daily Net Receipt figures in the Service Invoice.\nThe monthly Deliveries of –681,155 also match the “Nominated Delivered Total” in the Service Invoice.\nThis indicates that the two tables reconcile correctly for August 2001. However, the sign of the Current Month Imbalance is inconsistent.\nIn addition, the calculated Cumulative Imbalance for August 2001 is –403."
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Summary / Visualization

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Using the assumption area plus monthly detail on the Scenario1, Scenario2, and Scenario3 tabs, please add two charts per scenario: a hedge coverage stacked column chart with Month on the X-axis and stacked series (monthly contract MW), with an optional Short reference line; Y-axis in MW; show only the first year. Also add a portfolio cost line chart with Month on the X-axis and series Portfolio Weighted Avg (primary) and Monthly Price (reference), optionally including the 10-Yr/5-Yr/1-Yr Fixed Price lines; Y-axis in $/MWh; include all years.

{
  "files": [
    "26_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Planning and BudgetingSource task tags: Data Entry / Import · Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Compile the headcount for each department and update the Master sheet with the department-level totals. If a department has no headcount data, leave the corresponding cell blank.

{
  "files": [
    "27_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Planning and BudgetingSource task tags: Validation / Review · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Verify the accuracy of the department headcount summary by cross-checking each department’s total number of employees against its corresponding detailed roster sheet. Identify and correct any discrepancies—such as miscounts, missing entries, or remove any departments that no longer exist—and update the summary to reflect the accurate total headcount.

{
  "files": [
    "28_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Planning and BudgetingSource task tags: Structuring / Formatting · Calculation · Summary / Visualization

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Compile the existing budgets for each department and write each department’s summarized budget into column M of the Master sheet. If a person or department has no budget, treat the budget as 0. If a department does not exist, remove it from the sheet. After all departmental budgets are updated, calculate the total budget across all departments.

{
  "files": [
    "29_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Data Entry / Import · Cross-sheet/file Retrieval · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Based on the latest financials, complete all empty cells in the 'Total Wholesale Cash Flows excluding Prepays' table on Summary Average Term.

{
  "files": [
    "30_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Summary / Visualization

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Convert the employee rotation roster into a color-coded rotation plan Gantt chart on Rotation chart. The Rotation Chart tab should display each employee’s assignments by month for 2001 so the schedule is easy to review at a glance. Each color-coded rotation should also be labeled with the corresponding group name. To distinguish between the power and gas groups, add the suffix “-power” or “-gas” to the group names.

{
  "files": [
    "31_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPlanning and BudgetingSource task tags: Data Entry / Import · Structuring / Formatting · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Please prepare a summary of all groups and staffing as of March 1, including each group and business line’s supervisor and the number of Analysts (TT) and Associates (TT) on their team. The table should include four columns: Group, Line Supervisor (surname first), No. Analysts (TT), and No. Associates (TT).

{
  "files": [
    "32_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Gather Enron North America’s Mid Year 2001 performance across all departments into “All Originators by Value” sheet, showing each Originator, Commodity Team, Total, and % Total. Then add addiontal line to include the overall Total and %Total lines in the summary.

{
  "files": [
    "33_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationProcurement and SalesSource task tags: Calculation

Original English instruction · agent OS and apps are not declared.

Assume the following changes occur in the Jul–Dec 2002 market: Flat curve prices increase uniformly by $2/MWh; Peak 6x16 curve prices increase uniformly by $5/MWh; monthly contract volumes (Flat and Peak Total MWh) remain unchanged. Based on the 2002 table, calculate: (1) the total added value (mark-to-market change) for the combined Flat + Peak portfolio; and (2) what percentage of this added value comes from the Peak 6x16 contracts rather than the Flat contracts.

{
  "files": [],
  "text": "The total added value of the July–December 2002 portfolio is $1,989,600 (in absolute terms). Of this amount, approximately 27.9% (about 28%) comes from the Peak 6x16 contracts, with the remaining ~72.1% coming from the Flat contracts."
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Summarize the volume and dollar imbalances that exist between the various pipeline operators (Operators) and Transwestern.

{
  "files": [
    "35_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Calculation · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Prepare and complete the Exposure Table, which is used to track the company’s price exposure across multiple index points. The table should incorporate monthly contracted volumes and corresponding market prices for each pricing index, and calculate 55 days payables for each month using the formula: (prior month’s day count × Volume × Price) + (25 × Volume × Price).

{
  "files": [
    "36_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Append two columns to the end of the current explosure table: “Total” and “WA Price” (volumn-weighted average price averaged over the days).

{
  "files": [
    "37_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPricing and ValuationSource task tags: Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Using the discount rate assumptions in the table and each Shipper’s hurdle rate, term, volume, and annual rates, please complete and calculate the financial metrics for each Shipper—NPV, Actual Rate, Gross Value, cash flows by year and so on.

{
  "files": [
    "38_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Summary / Visualization · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Please create four charts in the “Graph Data Oct 01” spreadsheet, placing them in the empty area between rows 30 and 80. Please size them evenly so they look balanced and do not overlap. Chart 1 – Trend of Weekly Errors Rolling 60 Days Plot the total number of errors for each week from 7/30 to 10/1. Use a line chart, add a linear trendline, and show the data value at each point on the line. Chart 2 – Summary of Errors by Group for week of 10/1 Create a column chart where the X‑axis is Group (EIM, EGM, EEL, EA‑GAS, etc.). Include two data series: one for “# of errors” and one for “Ratio of Errors to Active Books.” Chart 3 – Trend of Book Creation Rolling 30 Day period For each Group, aggregate the Book creation counts by week. Use a stacked column chart with the X‑axis showing each week from 8/27 to 10/1, and the Y‑axis showing the number of books. Each Group should have its own color in the stack. Chart 4 – Breakout of Errors by Type per Week Rolling 60 Days Using the rows for each error type, create a stacked column chart by week. The X‑axis should be the weekly periods from 7/23 to 10/5, and the Y‑axis the number of errors. Each error type should be a different color, and the value for each segment should be shown as a data label on the stack.

{
  "files": [
    "39_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPlanning and Budgeting · ReportSource task tags: Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Audit the workbook and correct the formula errors in place so numbers calculate properly.

{
  "files": [
    "40_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationAccounts Payable and ReceivableSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared.

Treat “California Power Exchange” and “Pacific Gas and Electric (California Power Exchange) (a)” as a single combined CPX customer group. Reconstruct CPX’s full-cycle cash flows and net position across all sheets from November 2000 through June 2001, and determine whether, as of the certification date, the ISO is net receivable from CPX or net payable to CPX, along with a breakdown of the underlying reasons.

{
  "files": [],
  "text": "CPX as a whole is a net debtor to the ISO, with approximately $3.028 billion outstanding at period end."
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPricing and ValuationSource task tags: Calculation · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Using the Power Purchase Agreement (PPA) Cost and Savings Table for Cedar Braker, calculate and populate all relevant values in the summary table under the assumption of a 9.13% discount rate. The analysis should compute the Net Present Value (NPV) for each component — including the Old PPA, Market Stranded Investment (SI), New PPA, and PPA Savings — and compare the economic impact of the different scenarios.

{
  "files": [
    "42_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPricing and ValuationSource task tags: Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Complete the Cleburne Plant Damage Sensitivities table task Q1 – At Inception: First, based on the given assumptions in the table, fill in the Interest Rate Adjustment, Apache Savings, Adjusted Capacity Rate, Months in the Year, Plant Capacity, Yearly Capacity Payments, and Monthly Capacity Payments for each year from 1997 to 2019. Finally, provide the equity present value (XNPV5) as of 12/31/1996 and 12/31/2000.

{
  "files": [
    "43_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPlanning and BudgetingSource task tags: Summary / Visualization · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared.

On active deals vs headcount sheet, create two separate line charts to illustrate Enron Wholesale Services’ business and HR growth w/o Calgary from 1999 to 2001 : Chart 1: Active deals by month, showing separate lines for Natural Gas, Power, and Financial deals. Chart 2: Energy Operations headcount, comparing Adjusted Plan and Actual headcount over time.

{
  "files": [
    "44_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPlanning and BudgetingSource task tags: Data Entry / Import · Structuring / Formatting · Summary / Visualization

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Compile and populate the Summary sheet with Enron Energy Operations’ departmental annual headcount Plan vs Actual and then add a short analysis of the business drivers behind headcount growth for 1999–2001. Organize the commentary into three sections—Enron Americas, EGM, and EIM—and place it on the right-hand side of the Summary sheet.

{
  "files": [
    "45_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Complete the Financial Ratios section on the Balance Sheet. In addition, infer and populate the revaluation-related asset entries on rows 25 and 26.

{
  "files": [
    "46_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Complete the Income Statement (Purchase method) by calculating the Amortization of goodwill and the Amortization of extra depreciation, then update Income before income taxes. Based on that, compute the corresponding income tax, net income, and earnings per share on the Income Statement.

{
  "files": [
    "47_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Validation / Review · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Review the Income Statement for any mis-entered figures and correct them.

{
  "files": [
    "48_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Review the workbook and resolve the mismatched figures caused by formulas in the A1:K16 range that were entered without the necessary cell references. Identify the affected cells within that range and correct the formulas so the numbers reconcile.

{
  "files": [
    "49_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Complete the last few rows for “Inventory turnover” and the related metrics (average time in inventory, receivables turnover, and payables turnover) in line with the calculation relationships described in the sheet notes.

{
  "files": [
    "50_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Data Entry / Import · Structuring / Formatting

Original English instruction · agent OS and apps are not declared.

Identify the data in the chart and save it to a spreadsheet.

{
  "files": [
    "51_ref_0.xlsx"
  ],
  "text": ""
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Planning and BudgetingSource task tags: Data Entry / Import · Summary / Visualization · Calculation

Original English instruction · agent OS and apps are not declared.

Transcribe the content from the pdf/image into the Excel file (including the chart) and complete any missing formulas so the workbook is fully populated and the calculations are in place.

{
  "files": [
    "52_ref_0.xlsx"
  ],
  "text": ""
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPricing and Valuation · Trading and Risk Management · Procurement and SalesSource task tags: Calculation

Original English instruction · agent OS and apps are not declared.

For each month from December 2000 through April 2012, compute the difference between the Total Monthly Nominal Volume (CNG + TCO) and the Total Monthly Discounted Volume (CNG + TCO) for the long-term natural gas contract. Then compare the Average Daily Nominal Volume for CNG with the Average Daily Nominal Volume for TCO.

{
  "files": [],
  "text": "Total Monthly Nominal Volume (CNG+TCO): 75,100,712\nTotal Monthly Discounted Volume (CNG+TCO): 53,458,048\nTotal Monthly Nominal Volume (CNG+TCO) - Total Monthly Discounted Volume (CNG+TCO) = 21,642,663\nAverage Daily Nominal Volume (CNG): 11,409\nAverage Daily Nominal Volume (TCO): 5,619\nAverage Daily Nominal Volume (CNG) - Average Daily Nominal Volume (TCO)=5,790"
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Calculation · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Take the left-side raw pulp price data for BSCTMP, NBSK, and SBSK and add derived fields to calculate the correlations for BSCTMP vs. NBSK and BSCTMP vs. SBSK, identify the maximum and minimum values of Dollar difference and record the dates they occur, and report the standard deviation for each of BSCTMP, NBSK, and SBSK price series. Also calculates the annualized standard deviation (volatility) of the monthly log growth rates of BSCTMP, NBSK, SBSK, and present it in percentage terms to show relative volatility.

{
  "files": [
    "54_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Summary / Visualization

Original English instruction · agent OS and apps are not declared.

On the correl_graph sheet, create a time-series line chart comparing BSCTMP, NBSK, and SBSK prices to show how they move relative to each other. Use time on the x-axis to make their correlation visible.

{
  "files": [
    "55_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Structuring / Formatting · Calculation · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Reference the Summary sheet and restate the SUM-USD sheet on a USD basis. Update all figures and roll-ups in SUM-USD to reflect USD reporting consistent with the Summary.

{
  "files": [
    "56_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Calculation · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Review the YTD Recon sheet and compare Canada vs. Houston across Term, Cash, and the overall Total. Summarize the variance month-by-month and for the full year.

{
  "files": [
    "57_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

On the SUM-USD and Summary tabs, add a right-side data block that consolidates quarterly totals and the full-year total for each row. Please ensure every line is included.

{
  "files": [
    "58_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Update the TOTAL PHYSICAL GAS tab to mirror the layout on TOTAL US GAS. Specifically, insert a “% CHANGE FROM LAST 30 DAYS” column on the TOTAL PHYSICAL GAS worksheet.

{
  "files": [
    "59_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

On the All Natural Gas sheet, create an Excel Ctrl+T table and filter to show only the COUNTERPARTY entries highlighted in red.

{
  "files": [
    "60_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPricing and ValuationSource task tags: Data Entry / Import · Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Add two new entries to the peaker plant comps and then refresh the averages. Constellation (Illinois; Electric Power Daily, article 6/20/2000; operation 6/1/2001) with unspecified turbine type, 300 MW, total cost $130MM, and $433/kW; and TVA (Mississippi; Electric Power Daily, article 6/23/2000; operation 6/1/2002) using GE Gas Turbines, 340 MW, total cost $170MM, and $500/kW. After these are entered, update the overall average and the average excluding GE LM 6000 units.

{
  "files": [
    "61_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Calculation · Validation / Review

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

For EDF MAN, clear the 'Line of Credit Covering Initial Margin (except EDF Mann…)' and then recalculate any related figures that need to be synchronized. Leave all other content unchanged.

{
  "files": [
    "62_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Cross-sheet/file Retrieval · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Using RepIS-Qtrly as the base, please create the RepIS-Annual and RepIS-Qtrly YTD schedules, formatted consistent with the other spreadsheets in the workbook.

{
  "files": [
    "63_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Predictive ModelingSource task tags: Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Audit the workbook and correct the formula errors in place so numbers calculate properly.

{
  "files": [
    "64_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationProcurement and SalesSource task tags: Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Review the Inv & WC Value Adj summary tab and add the missing cross‑sheet data references to the other worksheets so the roll‑up pulls the correct figures. Return the updated file with those links in place.

{
  "files": [
    "65_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationProcurement and SalesSource task tags: Calculation · Structuring / Formatting · Financial Modeling

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Calculate the Interest Payment fpr enron and fill the correnponding cell.

{
  "files": [
    "66_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Structuring / Formatting · Calculation · Validation / Review · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Create a new 'by type_area' worksheet based on the Summary and the other tabs. It should present two separate tables summarized by Imbal Type; within each table, consolidate by area, include Volume, Value and Date, and calculate totals. Finally, confirm that the value and volume totals tie to the totals shown on the Summary.

{
  "files": [
    "67_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Cross-sheet/file Retrieval · Data Entry / Import

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Complete the Summary worksheet by entering the missing data and aligning it to the underlying sheets.

{
  "files": [
    "68_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Accounts Payable and ReceivableSource task tags: Structuring / Formatting · Validation / Review

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Please review the workbook and correct the naming errors in the CENT book entries—where a PHY tag should indicate a name with a PB prefix rather than FB. Once corrected in both Short Name and Long Name, set the corresponding row background to blue to flag the changes.

{
  "files": [
    "69_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Add a separate line item labeled “IT Infrastructure Allocation” under each ENA support department in the 2002 Plan and leave the amounts blank for now. This line will represent the planned IT infrastructure allocations ENA will pass through to the Business Units in 2002.

{
  "files": [
    "70_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Web Search · Data Entry / Import · Structuring / Formatting · Calculation · Summary / Visualization

Original English instruction · agent OS and apps are not declared.

Please list the top 10 individuals from the Forbes Global Billionaires Ranking for each year from 2019 to 2024 (including 2024). For each person, include their name, ranking, net worth, source of wealth, and age of that year end. Present the information in a table format. All your data should come from Forbes, which means the ages and net worth are calculated based on the publication date.\n\nPlease output an xlsx file with Sheet1 as the RawData sheet, with the column headers in order:\nYear, Rank, Name, Net Worth (in USD billions), Age, Source of Wealth\n\nDon't ask me any questions, just output the results according to the columns without omitting cells arbitrarily.\n\nBased on the RawData sheet, complete the following 4 subtasks:\n\nSubtask 1 - Sheet2_Wealth_Growth: Analyze wealth growth trends and year-over-year changes.\nColumns: Year, Name, Current Net Worth (USD billions), Annual Change (USD billions), Annual Change (%), Cumulative Growth (%), Growth Trend\nAnnual Change calculates difference from previous year (N/A for first appearance, 2 decimals). Cumulative Growth (%) is relative to first year on list (2 decimals). Growth Trend: \"Rapid Growth\" if Annual Change % ≥ 20%, \"Steady Growth\" if ≥ 5%, \"Stable\" if -5% to 5%, \"Moderate Decline\" if -20% to -5%, \"Significant Decline\" if < -20%, \"New Entry\" if no previous year data. Net worth values to 2 decimals.\n\nSubtask 2 - Sheet3_Industry_Concentration: Aggregate by industry sector and analyze market concentration.\nColumns: Year, Industry, Billionaire Count, Total Wealth (USD billions), Average Wealth (USD billions), Market Share (%), Concentration Level\nIndustry standardization: Amazon → E-Commerce, Microsoft → Technology-Software, Tesla/SpaceX → Technology-Automotive/Aerospace, Facebook/Meta → Technology-Social Media, Google/Alphabet → Technology-Internet Services, Oracle → Technology-Enterprise Software, LVMH → Luxury Goods, Berkshire Hathaway → Diversified Holdings, Walmart → Retail, Zara → Fashion Retail, Bloomberg → Financial Services, Telecom → Telecommunications, Reliance → Conglomerate. Market Share (%) = (Industry Total / Year Total) × 100 (2 decimals). Average Wealth = Total / Count (2 decimals). Concentration Level: \"Monopolistic\" if count = 1, \"Highly Concentrated\" if count = 2, \"Concentrated\" if count ≤ 4, \"Diversified\" if count > 4. Sort by Year ascending, then Total Wealth descending within each year. Billionaire Count is integer.\n\nSubtask 3 - Sheet4_Ranking_Stability: Track individual ranking stability and volatility over time.\nColumns: Name, Years on List, First Year, Last Year, Best Rank, Worst Rank, Average Rank, Rank Volatility, Ranking Trend, Stability Rating\nYears on List = count of appearances (integer). Best Rank = minimum rank number (integer). Worst Rank = maximum rank number (integer). Average Rank = mean of all ranks (2 decimals). Rank Volatility = standard deviation of ranks, 0.00 if only one year (2 decimals). Ranking Trend: \"Rising\" if average annual rank change ≤ -1, \"Slightly Rising\" if ≤ -0.2, \"Stable\" if -0.2 to 0.2, \"Slightly Falling\" if ≤ 1, \"Falling\" if > 1 (rank change calculated as (last rank - first rank) / years span). Stability Rating: \"Highly Stable\" if years ≥ 6 and volatility ≤ 1.5, \"Stable\" if years ≥ 5 and volatility ≤ 2.5, \"Moderately Stable\" if years ≥ 4 and volatility ≤ 3.5, \"Volatile\" if years ≥ 3, \"Insufficient Data\" otherwise. Sort by Years on List descending, then by order of first appearance in RawData for ties. First Year, Last Year, Best Rank, Worst Rank are integers.\n\nSubtask 4 - Sheet5_Pivot_Summary: Create pivot summary showing top 3 and aggregate statistics by year.\nColumns: Year, Rank 1 Name, Rank 1 Wealth (USD billions), Rank 2 Name, Rank 2 Wealth (USD billions), Rank 3 Name, Rank 3 Wealth (USD billions), Total Top 10 Wealth (USD billions), Average Top 10 Wealth (USD billions), Median Age, Wealth Gap (Rank 1 vs Rank 10), Top 3 Concentration (%)\nTotal Top 10 Wealth = sum of all 10 billionaires (2 decimals). Average Top 10 Wealth = Total / 10 (2 decimals). Median Age = median of 10 ages (integer). Wealth Gap = Rank 1 wealth - Rank 10 wealth (2 decimals). Top 3 Concentration (%) = (Sum of Top 3 / Total Top 10) × 100 (2 decimals). All wealth values to 2 decimals. Sort by Year ascending.

{
  "files": [
    "71_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Data Entry / Import · Summary / Visualization · Calculation · Structuring / Formatting

Original English instruction · agent OS and apps are not declared.

Transcribe the content from the pdf/image into the Excel file (including the chart) and complete any missing formulas so the workbook is fully populated and the calculations are in place.

{
  "files": [
    "72_ref_0.xlsx"
  ],
  "text": ""
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationAccounts Payable and ReceivableSource task tags: Cross-sheet/file Retrieval · Calculation

Original English instruction · agent OS and apps are not declared.

For account type H and the MCC description: Eating Places and Restaurants, what would be the average fee that the card scheme GlobalCard would charge for a transaction value of 10 EUR? Provide the answer in EUR and 6 decimals. Answer must be just a number expressed in EUR rounded to 6 decimals. If a question does not have a relevant or applicable answer for the task, please respond with 'Not Applicable'

{
  "files": [],
  "text": "0.123217"
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge",
    "source_csv_json_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input.",
    "The preprocessing chain handles PDF, Markdown, Word, Excel and images, with no CSV/JSON handler. The content builder attaches source content through the first Excel path; its general-side builder covers reference/output, not source files. IDs 17, 18 and 73 contain CSV/JSON inputs. Do not claim the current default judge directly inspects those structured inputs. Keep reference text and original source links; assess coverage before any benchmark run."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    },
    {
      "issue_id": "FINCH-E03",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L113-L126",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/build_prompt/content_builder/content_builder.py#L205-L261"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationAsset ManagementSource task tags: Structuring / Formatting · Validation / Review · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Rmove the Physical Location column from the Detail by Turbine sheet, and keep the Physical Location values in the Summary by Status sheet exactly as they are. Ensure that deleting columns does not affect the content of the current sheet or other sheets.

{
  "files": [
    "74_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationAsset ManagementSource task tags: Summary / Visualization · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

On the Summary by Status sheet, add a cumulative curve chart ($MM) for “Scheduled vs Cancellation Payments” using monthly data from Dec-98 through Dec-02. Plot the cumulative series for Scheduled and Cancellation across Committed, Tentative, and Available, and include Total Scheduled and Total Cancellation (Total Cancellation represents total exposure). Add a blue vertical line at Nov-2001 and annotate that month’s cumulative cancellation cost and cumulative paid amounts in blue, with the difference shown in red as “Incremental Cancellation Cost” (e.g., $35MM).

{
  "files": [
    "75_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Reformat the table by bolding the titles and inserting row borders between the different departments to better delineate each section.

{
  "files": [
    "76_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Calculation · Validation / Review

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Calculate the headcount for each of the three groups in the worksheet and their respective percentages of the total, and confirm in the table that the three percentages sum to 100%.

{
  "files": [
    "77_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Planning and Budgeting · ReportSource task tags: Cross-sheet/file Retrieval · Validation / Review

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Review the summary tab against each of the individual sheets and reconcile any inconsistencies. Where the summary differs from the sub-sheets, update the summary to reflect the sub-sheet values so the workbook is internally consistent.

{
  "files": [
    "78_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Planning and Budgeting · ReportSource task tags: Cross-sheet/file Retrieval · Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Consolidate data from each business unit worksheet into a new ETS worksheet, and add three columns: Total, Consol, and Enron: Total: Sum of all business units Consol: Consolidation adjustments to eliminate intercompany transactions from Citrus and NBP Enron: Final consolidated company figures (Total + Consol)

{
  "files": [
    "79_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Planning and Budgeting · ReportSource task tags: Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Restore the missing formulas so the totals and Net Income rollups calculate correctly, converting any hard-coded cells back to formula-driven values. This should bring the sheet back to a consistent, automated state for the 2001 functional income schedule.

{
  "files": [
    "80_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Planning and Budgeting · ReportSource task tags: Data Entry / Import · Structuring / Formatting · Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared.

Key the table into Excel on an "ETS" spreadsheet and reinstate the missing formulas on the ETS sheet. Also update the I1 header to “Other” and ensure the totals and Net Income roll-ups (columns J-L around row 225 and the Net Income row at 228) calculate correctly.

{
  "files": [
    "81_ref_0.xlsx"
  ],
  "text": ""
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

On the 'simplecorr' sheet, create a table whose column headers mirror the paper types in the left-side table, with a leftmost header labeled 'Correlation.' Using the NBSK pulp price from the left table and each paper’s monthly price series, compute and populate the correlation of NBSK to each paper’s price.

{
  "files": [
    "82_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

On the simplecorr sheet, create a table that mirrors the header from the table to the left to list the various paper grades. Set the left-side header to 1–6 months Lag and Lead, and then, using the NBSK pulp price from the left table together with the monthly price tables for each paper grade, calculate and populate the 1–6‑month Lag and 1-7 month Lead relationships between NBSK and each paper price.

{
  "files": [
    "83_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Calculation · Financial Modeling · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

On the Regression sheet, run a linear regression to examine the relationship between "Std. No. 4, 83–85 Brt Xerog." and NBSK chg, with NBSK as the dependent variable. Summarize the results in a single table with the top header listing "Std. No. 4, 83–85 Brt Xerog.", and NBSK, and the left header listing mn, se, r2, f, and ssreg.

{
  "files": [
    "84_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Structuring / Formatting · Calculation · Financial Modeling

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Replicate the linear regression analysis we previously ran on the No. 4 Xero vs NBSK chg relationship, extending the same approach to the other paper price series and summarizing their impact on NBSK in a table. Use the same column headers as the No. 4 Xero–NBSK chg regression table.

{
  "files": [
    "85_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Data Entry / Import · Calculation · Validation / Review

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Complete the asset allocation schedule using the provided asset detail data by filling in any blank items. First calculate Total Equity and Total (total assets). Then compute Cash %, Equity %, and Fixed Income %, round each to two decimals, populate the respective percentage fields, and confirm the three sum to approximately 100%.

{
  "files": [
    "86_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Web Search · Data Entry / Import · Calculation · Structuring / Formatting

Original English instruction · agent OS and apps are not declared.

Could I obtain monthly data from January 2025 to May 2025 (including January 2025 and May 2025) for NASDAQ, NYSE, Shanghai Stock Exchange, Shenzhen Stock Exchange, and Hong Kong Exchanges and Clearing?\n\nPlease output an xlsx file with Sheet1 as the RawData sheet, with the following column names in order: Exchange Name, Statistical Month, Total Trading Value (USD millions), Total Number of Listed Companies, Domestic Market Capitalization (USD millions), Index Levels, Average Daily Trading Value (USD millions)\nFor the statistical month, keep the format as January 2025.\nFor total trading value and domestic market capitalization, both of the terms refer to equity.\nDon't ask me any questions, just output the results according to the columns without omitting cells arbitrarily.\n\nBased on the RawData sheet, complete the following 4 subtasks:\n\nSubtask 1 - Sheet2_KPIs: Calculate 4 KPI indicators for each exchange-month.\nColumns: Exchange Name, Statistical Month, Trading Value per Company (USD millions), Market Cap per Company (USD millions), Daily Trading Intensity (USD millions), Market Efficiency Ratio\nTrading Value per Company = Total Trading Value / Total Number of Listed Companies (2 decimals). Market Cap per Company = Domestic Market Capitalization / Total Number of Listed Companies (2 decimals). Daily Trading Intensity = Average Daily Trading Value (2 decimals). Market Efficiency Ratio = Total Trading Value / Domestic Market Capitalization (4 decimals).\n\nSubtask 2 - Sheet3_MonthlyRanking: Rank exchanges by trading value within each month independently.\nColumns: Exchange Name, Statistical Month, Total Trading Value (USD millions), Monthly Rank, Is Top 3\nEach month should have ranks 1-5, not global ranking. Is Top 3: \"Yes\" or \"No\". Sort by month chronologically (January to May), then by Monthly Rank ascending within each month.\n\nSubtask 3 - Sheet4_GrowthAnalysis: Calculate month-over-month growth rates for February-May only.\nColumns: Exchange Name, Current Month, Previous Month, Trading Value Growth (%), Market Cap Growth (%)\nGrowth = (Current - Previous) / Previous × 100 (2 decimals). Sort by Exchange Name alphabetically, then by month chronologically (February to May) within each exchange.\n\nSubtask 4 - Sheet5_Summary: Statistical summary for each exchange across 5 months.\nColumns: Exchange Name, Total Trading Value (USD millions), Avg Monthly Trading Value (USD millions), Max Monthly Trading Value (USD millions), Min Monthly Trading Value (USD millions), Std Dev of Trading Value (USD millions), Coefficient of Variation (%), Avg Market Capitalization (USD millions), Avg Number of Companies\nCoefficient of Variation = Std Dev / Avg × 100. Sort by Total Trading Value descending.

{
  "files": [
    "87_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Web Search · Data Entry / Import · Structuring / Formatting · Calculation · Summary / Visualization

Original English instruction · agent OS and apps are not declared.

Need to analyze the trend of the U.S. federal government spending and deficit before and after the pandemic. Please provide me with the following data through fiscal years 2015-2024: the federal Budget (trillion), the federal spending (trillion), the federal deficit (trillion), the national debt (trillion), and the net interest cost on the gross federal debt (trillion).\n\nPlease output an xlsx file with Sheet1 as the RawData sheet, with the following column names in order:\nFiscal Year, Federal Budget, Federal Spending, Federal Deficit, National Debt, Net Interest Cost \n\nUnder the Fiscal Year, state the statistics like FY2015, FY2016.\n\nDon't ask me any questions, just output the results according to the columns without omitting cells arbitrarily.\n\nBased on the RawData sheet, complete the following 4 subtasks:\n\nSubtask 1 - Sheet2_Pandemic_Comparison: Compare pre-pandemic (FY2015-2019) vs post-pandemic (FY2020-2024) statistics.\nColumns: Period, Avg Budget, Avg Spending, Avg Deficit, Avg National Debt, Avg Interest Cost, Total Years\nAll averages with 2 decimals.\n\nSubtask 2 - Sheet3_Fiscal_Indicators: Calculate 5 fiscal health metrics for each year.\nColumns: Fiscal Year, Budget Deficit Rate, Debt-to-Budget Ratio, Interest-to-Spending Ratio, Spending Efficiency, Debt Service Burden\nBudget Deficit Rate = Deficit / Budget × 100 (2 decimals). Debt-to-Budget Ratio = Debt / Budget (2 decimals). Interest-to-Spending Ratio = Interest / Spending × 100 (2 decimals). Spending Efficiency = Budget / Spending × 100 (2 decimals). Debt Service Burden = Interest / Debt × 100 (4 decimals).\n\nSubtask 3 - Sheet4_Growth_Analysis: Calculate year-over-year growth rates for FY2016-FY2024.\nColumns: Fiscal Year, Budget Growth, Spending Growth, Deficit Growth, Debt Growth, Interest Cost Growth\nAll growth rates = (Current - Previous) / Previous × 100 (2 decimals).\n\nSubtask 4 - Sheet5_Summary: Overall 10-year statistics.\nColumns: Total Budget, Total Spending, Total Deficit, Avg Annual Debt, Total Interest Paid, Max Single-Year Deficit, Max Deficit Year, Overall Deficit Rate, Avg Interest Rate, Debt Growth Rate\nOverall Deficit Rate = Total Deficit / Total Budget × 100 (2 decimals). Avg Interest Rate = Total Interest / Avg Debt × 100 (4 decimals). Debt Growth Rate = (FY2024 Debt - FY2015 Debt) / FY2015 Debt × 100 (2 decimals).

{
  "files": [
    "88_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Web Search · Data Entry / Import · Structuring / Formatting · Calculation · Validation / Review

Original English instruction · agent OS and apps are not declared.

I am working on a tracking report regarding the biotechnology and pharmaceutical companies that went public on NASDAQ in 2024, namely: CG Oncology, Zenas BioPharma, Upstream Bio, MBX Biosciences, and Metagenomi. I want to start by organizing some data. Please help me find out the listing date (as in yyyy/mm/dd), listing board(full name), initial offering price, total funds raised in 2024. Additionally, I need the revenue, net income attributable to shareholders and R&D expenses disclosed in their 2024 annual reports for these companies. All amounts should be in United States dollars, numerical only, and retained to two decimals. If no relevant data can be found, please fill in with N/A.\n\n[Data Time Baseline] Please retrieve data based on each company's 2024 10-K annual report (versions published before March 31, 2025). For 'Total Funds Raised in 2024', only count IPO initial offering proceeds (including over-allotment), excluding subsequent follow-on offerings.\n\nPlease output an xlsx file with Sheet1 as the RawData sheet, with the following column names in sequence:\nCompany Name, Listing board, Bloomberg ticker, Listing Date, Initial Offering Price(per share), Total Funds Raised in 2024(million), Revenue in 2024(million), Net Income Attributable to Shareholders(million), R&D Expenses(million)\n\nDon't ask me any questions, just output the results according to the column without omitting cells arbitrarily.\n\nBased on the RawData sheet, complete the following 4 subtasks:\n\nSubtask 1 - Sheet2_Financial_Metrics: Calculate financial performance indicators.\nColumns: Company Name, Revenue in 2024(million), Net Income Attributable to Shareholders(million), R&D Expenses(million), R&D Intensity(%), Net Profit Margin(%), Capital Efficiency Ratio, Burn Rate Category\nR&D Intensity = R&D Expenses / Revenue x 100 (2 decimals). Net Profit Margin = Net Income / Revenue x 100 (2 decimals). Capital Efficiency Ratio = Revenue / Total Funds Raised in 2024 (4 decimals). Burn Rate Category: \"Profitable\" if Net Income >= 0, \"Moderate Burn\" if -100 < Net Income < 0, \"High Burn\" if Net Income <= -100, \"Unknown\" if data missing.\n\nSubtask 2 - Sheet3_Data_Quality: Validate data completeness.\nColumns: Company Name, Has Total Funds Raised in 2024, Has Revenue in 2024, Has Net Income Attributable to Shareholders, Has R&D Expenses, Data Completeness(%), Quality Rating, Missing Fields Count\nFor each field: \"Yes\" if data exists, \"No\" if N/A or empty. Data Completeness(%) = (Number of fields with data / 4) x 100 (2 decimals). Quality Rating: \"Excellent\" if 100%, \"Good\" if >=75%, \"Fair\" if >=50%, \"Poor\" if <50%.\n\nSubtask 3 - Sheet4_IPO_Timeline: Reorganize by chronological order.\nColumns: Company Name, Bloomberg Ticker, Listing Date, Initial Offering Price(per share), Total Funds Raised in 2024(million), IPO Quarter, Days Since First IPO, IPO Sequence\nSort by Listing Date (earliest first). IPO Quarter: Q1/Q2/Q3/Q4 based on month. Days Since First IPO: days elapsed since CG Oncology's IPO on 2024/01/25 (integer). IPO Sequence: chronological order number 1, 2, 3... (integer).\n\nSubtask 4 - Sheet5_Comparative: Conduct comparative analysis with ranking.\nColumns: Company Name, Total Funds Raised in 2024(million), R&D Expenses(million), Funds Raised Rank, R&D Expense Rank, Market Position, Investment Scale, Composite Score\nRanks: 1=highest among valid values, \"N/A\" if data missing. Market Position: \"Premium Tier\" if IPO price >= $19, \"Mid-High Tier\" if >= $17, \"Standard Tier\" otherwise. Investment Scale: \"Large Scale\" if funds >= $400M, \"Mid Scale\" if >= $250M, \"Unknown\" otherwise. Composite Score: (Funds/437)x50 + (R&D/139.14)x50 (2 decimals), \"N/A\" if both values missing.

{
  "files": [
    "89_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk Management · Investment: Credit · Pricing and ValuationSource task tags: Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Add a top border to all values in the Summary tab that are calculated as the sum of other rows.

{
  "files": [
    "90_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPricing and Valuation · Trading and Risk ManagementSource task tags: Calculation · Financial Modeling

Original English instruction · agent OS and apps are not declared.

On December 11, 2001, a trading firm enters into a long Cinergy–PJM‑W basis spread at the respective mid prices: it goes long 100 MW Jul–Aug 2002 baseload at Cinergy and short 100 MW Jul–Aug 2002 baseload at PJM‑W. Two weeks later, market prices move to: Cinergy Jul–Aug 2002 quoted at 49.00/49.25, and PJM‑W Jul–Aug 2002 quoted at 54.00/54.25. You are required first to use the original quotes to calculate the mid prices at both hubs and the initial spread, then use the new quotes to calculate the new mid prices and the new spread. Next, compute the mark‑to‑market P&L over the two‑week period for this 100 MW spread position (showing the P&L on the Cinergy leg, the PJM‑W leg, and the combined position). Finally, based on the change in the spread, determine whether Cinergy has strengthened relative to PJM‑W or vice versa, and discuss, assuming this spread was intended to hedge the risk of “selling power in PJM‑W and buying power in Cinergy,” whether the hedge performed effectively and why.

{
  "files": [],
  "text": "The 100 MW long Cinergy–PJM-W spread earned approximately $383k over the two-week period. Cinergy strengthened relative to PJM-W (or equivalently, PJM-W weakened). From a basis-risk perspective, the P&L of this long spread effectively offsets the margin impact of “selling power in PJM-W and buying power in Cinergy.” In other words, it locks in the future cross-hub price difference at the entry-level (around 7.575 $/MWh), making the hedge theoretically very effective."
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPricing and Valuation · Trading and Risk ManagementSource task tags: Calculation · Financial Modeling

Original English instruction · agent OS and apps are not declared.

At the Cinergy hub, a retail power company wants to hedge a 50 MW baseload position for the year 2002. It has two alternatives: Strategy A is to buy 50 MW of the Jan–Dec 2002 annual strip at the mid price; Strategy B is to buy, also at mid prices, 50 MW of each of the seven seasonal strips listed above, so that together they cover the entire year. You are asked first to compute the mid price of each seasonal strip and of the annual strip. Then, for Strategy B, calculate for each strip the total energy volume (in MWh) and the total cost, and use these to derive the volume‑weighted average hedge price for the whole year under Strategy B. Finally, compare this average price with the annual‑strip mid price under Strategy A, determine which strategy has the lower nominal hedging cost and by how many $/MWh, and, taking into account liquidity, execution complexity, and the shape of the seasonal load profile, explain which strategy you would recommend and why.

{
  "files": [],
  "text": "Strategy A: Buy the Cinergy Jan–Dec 2002 annual strip at the mid price (PA = 30.375 $/MWh ≈ 30.38 $/MWh). Strategy B: Use the seven seasonal strips to form the full-year exposure, giving a volume-weighted average price PB ≈ 30.44 $/MWh. Strategy A is cheaper by about 0.065 $/MWh (≈ 28,000 $/year for a 50 MW baseload). Without other considerations, I would recommend Strategy A."
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationProcurement and SalesSource task tags: Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Complete both the Flat and Peak tables by using the provided Direct Sales contract data and monthly Curve Prices. For each deal, calculate the corresponding monthly MWh. Populate the monthly MWh values under the appropriate counterparty rows, compute the Total MWh for each month, and then calculate the monthly Value using Total MWh × Curve Price. The final deliverable is a complete, accurate, and audit-ready monthly summary of MWh and Value for both Flat and Peak products

{
  "files": [
    "93_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPricing and Valuation · Trading and Risk ManagementSource task tags: Calculation

Original English instruction · agent OS and apps are not declared.

For September–December 2001 and September–December 2002, compute the monthly ratios of ERCOT peak power prices to NYMEX natural gas prices. Then calculate, for each period, the average ratio and the standard deviation (as a measure of volatility). Based on these results, determine in which period power prices were more “stable” relative to gas prices (using lower ratio volatility as the criterion).

{
  "files": [],
  "text": "For September–December 2001, the average ratio of power price to gas price is approximately 9.44, with a standard deviation of about 1.52.\nFor September–December 2002, the average ratio is approximately 8.43, with a standard deviation of about 0.59.\nSince the standard deviation in 2002 is smaller, ERCOT peak power prices were relatively more “stable” against NYMEX natural gas prices in Sep–Dec 2002 (based on the criterion of lower ratio volatility)."
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · ReportSource task tags: Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared.

From the PJM dataset for January 1997 through April 2000, identify all months that simultaneously satisfy the following three conditions: Peak Demand > 45,000 Megawatt Daily Pricing Average > 30 Henry Hub gas price > 2.5

{
  "files": [],
  "text": "Aug 1999"
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Financial Modeling · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Run the holding-period return analysis for ChinaBond Export‑Import Bank debt assuming a 0.5‑year hold, comparing the 1‑year vs. 4‑year maturities. Use the scenario where, over the next six months, the 1Y yield shifts up by 1bp and the longer tenor shifts up by 6bp, with current coupons set at 1.25% and 1.75%, respectively.

{
  "files": [
    "96_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Structuring / Formatting · Calculation · Financial Modeling

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Finalize the stock-selection model by completing the cross-sectional ranking formulas in columns K/L/M on Sheet1, with M aligned to the one-year Sharpe ranking and K/L aligned to their respective factor columns per the headers. Once the formulas are in place, use the results in column O to populate Sheet2 with the final selections—two columns (stock code and weight)—showing only those with weights greater than zero.

{
  "files": [
    "97_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Data Entry / Import · Web Search

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Use publicly available market/financial data to populate Sheet1 columns F, G, and H—namely the 12‑month dividend yield, last quarter YoY profit growth, and current quarter YoY profit growth—for each constituent security. No other changes are required to the workbook.

{
  "files": [
    "98_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Based on the Canada – Non-Commercial roster, prepare a headcount summary by functional area, showing how many employees fall into Group 1, Group 2, and Group 3 in each department.

{
  "files": [
    "99_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Structuring / Formatting · Calculation · Validation / Review

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Update the Canada Non-Commercial functional distribution table, ensuring the table remains aligned with the latest staffing data in Sheet2. Additionally, add two columns — Total and %Total — to display the total headcount and percentage for each group.

{
  "files": [
    "100_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Standardize all date entries in the spreadsheet to the 01-FEB-2002 format (dd-MMM-yyyy). Month headers should read consistently as 01-JAN-2002, 01-FEB-2002, etc. Pay attention to case sensitivity.

{
  "files": [
    "101_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Structuring / Formatting · Calculation · Validation / Review

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Create a new Transfer worksheet that summarizes LONGS, SHORT, and NET by delivery month, with Total Longs, Total Shorts, and Total Net at the bottom; the month list should follow the Z, F1… sequence covering all months present, and show 0 for months with no data. In Access Trades, add a side display for TOTAL LONG and TOTAL SHORT with two Notes columns to confirm these tie to the Transfer totals (“ok” if consistent, otherwise note the variance), and return the updated workbook including the Transfer summary and the Access Trades validation block.

{
  "files": [
    "102_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPricing and ValuationSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

On Sheet4, create two monthly, time-aligned summaries: 'Peoples Deal Value' should sum the Mid Value by month from the 'Peoples Baseload Sale' and 'Transport' sheets (same-month totals), with months without data shown as 0 and a grand total at the end; 'PERC Deal Value' should sum the Mid Value by month from the 'PERC' sheet, with months without data shown as 0. Ensure the month sets for both tables are aligned and sorted chronologically.

{
  "files": [
    "103_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Predictive ModelingSource task tags: Structuring / Formatting · Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Insert a 'Total Available' row and a 'Total Allocated' row, then complete the reconciliation line to verify that 'Total Available' and 'Total Allocated' are consistent.

{
  "files": [
    "104_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Predictive ModelingSource task tags: Data Entry / Import · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Add the 2/11/2000 column on the Feb 00 tab by mirroring the 2/10 data, with the following adjustments: Houston Pipe Line should be 6,000 (instead of 9,000), and for Huntsville and Woodlands use 1,000 and 5,000 respectively. All other items remain unchanged from 2/10.

{
  "files": [
    "105_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

On Sheet3, append the month-over-month percentage changes for BSCTMP and SBSK, then compute the Pearson correlation between their monthly price change rates.

{
  "files": [
    "106_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

On Sheet3, insert on the left side of the sheet a results table with a top-header showing the series names “BSCTMP” and “SBSK,” and two subsections with left-hand labels “1–6 month lag” and “1–6 month lead. In the 1–6 month lag section, quantify the timing relationship by calculating Pearson correlations between SBSK and BSCTMP shifted 1–6 months later and entering the resulting coefficients under the headings “1 month lag” through “6 month lag.” In the 1–6 month lead section, calculate correlations between BSCTMP and SBSK shifted 1–6 months later and entering the coefficients under the headings “1 month lead” through “6 month lead.

{
  "files": [
    "107_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReport · Planning and BudgetingSource task tags: Cross-sheet/file Retrieval · Data Entry / Import

Original English instruction · agent OS and apps are not declared.

Update the data in first source file based on the second source file.

{
  "files": [
    "108_ref_0.xlsx"
  ],
  "text": ""
}

{
  "review_flags": [
    "first_source_excel_only"
  ],
  "review_notes": [
    "The content builder selects only the first Excel file for each source/reference/output side. IDs 3 and 108 each have two source XLSX inputs; ID 108 explicitly requests use of the second. Excel preprocessing recalculates files but returns no textual evidence entries. Record source-visibility limits; verify all required inputs appear in any future judge payload. This does not establish task infeasibility or a measured grading error."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E01",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/build_prompt/content_builder/content_builder.py#L171-L261",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/build_prompt/content_builder/utils.py#L64-L75",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_excel.py#L305-L332"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Planning and BudgetingSource task tags: Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Calculate the total FTE percentage by region and by business line, and roll these up into a consolidated summary. This should include totals across EWS, EES, EGM, M&A, and EGA for each region (US Energy, Canada, Federal), with an overall summary of the combined FTE percentages.

{
  "files": [
    "109_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPlanning and BudgetingSource task tags: Calculation · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Using the daily PGE/SCG spot midpoints, the TW base price, the fuel‑loss factor, the fixed commodity cost (0.0246), and the day’s delivered volumes to PGE and SCG, please calculate Adj TW Per, the net unit spreads at PGE and SCG, the theoretical profit (spread × volume), the daily total profit, and the Astra/TW 70%/30% profit split, and roll these up to a full‑month summary. Among these values, keep four decimal places for Adj TW Per, the net unit spreads at PGE and SCG, and keep two decimal places for all other figures.

{
  "files": [
    "110_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPlanning and BudgetingSource task tags: Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Recompute and update the entire workbook under the new operating assumptions, without changing the original price and volume inputs (PGE/SCG Midpoint, TW benchmark price, and commodity cost at 0.0246). Reduce the TW→PGE fuel loss factor to 4.0% and change the profit split to Astra 60% / TW 40%, which better reflects joint sharing of market and scheduling risk.

{
  "files": [
    "111_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Accounts Payable and ReceivableSource task tags: Cross-sheet/file Retrieval · Data Entry / Import · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

For each record, use the Frequency to place the Rent amount into the corresponding Total Weekly, Total Biweekly, or Total Monthly column, and leave the non-applicable total columns blank. Then roll up the Sub-Total for each frequency and estimate the Annual Total.

{
  "files": [
    "112_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · Planning and BudgetingSource task tags: Data Entry / Import · Calculation · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Update the Active Deals vs Headcount sheet with the monthly Active Deals and headcount (HC), and capture the growth amount and the percentage increase from Sep 99 to Mar 01.

{
  "files": [
    "113_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPricing and Valuation · Trading and Risk ManagementSource task tags: Data Entry / Import · Cross-sheet/file Retrieval · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Using the trades as the source, identify and fill in the applicable asset and liability items for ENA Corp and ECT Inv, then roll them up into a summary. Please ensure the entries reflect only the portions relevant to each entity.

{
  "files": [
    "114_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Web Search · Data Entry / Import · Structuring / Formatting · Calculation · Summary / Visualization

Original English instruction · agent OS and apps are not declared.

Please track and compile the average monthly house prices for England, Scotland, Wales and Northern Ireland, for the period from January 2022 to April 2025 (including January 2022 and April 2025). Please cite all the statistics from government websites.\n\n[Data Source Note] Please use the UK House Price Index (HPI) data published on GOV.UK / HM Land Registry or ONS official sources. Note: HPI data is periodically revised; data from different publication dates is acceptable as long as sourced from official government websites. For data from February 2024 onwards, rounded values (to nearest thousand) from ONS bulletins are acceptable.\n\nPlease output an xlsx file with Sheet1 as the RawData sheet, with the following column names in order: Month, House price in England(Pounds), House price in Scotland(Pounds), House price in Wales(Pounds), House price in Northern Ireland(Pounds)\nFor the column 'Month', list the month in the format 2022/01, 2022/02.\n\nDon't ask me any questions, just output the results according to the columns without omitting cells arbitrarily.\n\nBased on the RawData sheet, complete the following 4 subtasks:\n\nSubtask 1 - Sheet2_Quarterly_Aggregation: Convert monthly data to quarterly averages by region.\nColumns: Quarter, England Avg Price, Scotland Avg Price, Wales Avg Price, Northern Ireland Avg Price\nQuarter format: 2022-Q1, 2022-Q2, etc. Each price value with 2 decimals. Sort chronologically from 2022-Q1 to 2025-Q2.\n\nSubtask 2 - Sheet3_Price_Changes: Calculate month-over-month price changes.\nColumns: Month, England MoM Change (%), Scotland MoM Change (%), Wales MoM Change (%), Northern Ireland MoM Change (%)\nAll percentages with 2 decimals. Start from 2022/02 (comparing to 2022/01).\n\nSubtask 3 - Sheet4_Price_Classification: Classify each region's price level for each month.\nColumns: Month, England Category, Scotland Category, Wales Category, Northern Ireland Category\nClassification rules: \"Very High\" if >= 310,000 (England), >= 200,000 (Scotland), >= 225,000 (Wales), >= 190,000 (Northern Ireland); \"High\" if >= 300,000 / >= 190,000 / >= 215,000 / >= 180,000; \"Medium\" if >= 290,000 / >= 180,000 / >= 205,000 / >= 170,000; \"Low\" otherwise.\n\nSubtask 4 - Sheet5_Summary_Statistics: Calculate summary statistics for each region.\nColumns: Region, Max Price, Max Price Month, Min Price, Min Price Month, Overall Change (%), Peak-to-Current Change (%), Volatility Score\nOverall Change = (Price in 2025/04 - Price in 2022/01) / Price in 2022/01 x 100 (2 decimals). Peak-to-Current Change = (Price in 2025/04 - Max Price) / Max Price x 100 (2 decimals). Volatility Score = STDEV.P (population standard deviation) of all monthly prices / AVERAGE x 100 (2 decimals). Max/Min Price with 2 decimals.

{
  "files": [
    "115_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPricing and Valuation · Trading and Risk ManagementSource task tags: Calculation

Original English instruction · agent OS and apps are not declared.

Compare the month-to-month movement direction of ERCOT peak power prices and NYMEX natural gas prices in 2002 (i.e., whether each month is higher or lower than the previous month). Based on the 12 monthly changes in 2002, determine how many months the two prices moved in the same direction and how many months they moved in opposite directions.

{
  "files": [],
  "text": "Same direction: 8 months\nOpposite direction: 2 months\nAdditionally, there are 2 months where one series is unchanged while the other moves, so they fall into neither category."
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Standardize the font across all data entries so they ues a single, consistent typeface throughout. No content changes are required; this is a formatting cleanup only.

{
  "files": [
    "117_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared.

What is the TW EOL charge for 2002? Pls just provide the amount.

{
  "files": [],
  "text": "113291"
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Calculation

Original English instruction · agent OS and apps are not declared.

How many plants are recorded in the spreadsheet?

{
  "files": [],
  "text": "85"
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Audit the workbook and correct the formula errors in place so numbers calculate properly.

{
  "files": [
    "120_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPlanning and Budgeting · Trading and Risk ManagementSource task tags: Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Audit the Deal Sheet in the workbook and correct the formula errors in place so numbers calculate properly.

{
  "files": [
    "121_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Summary / Visualization · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Create a new sheet named “Exp by Fun Gen Support Chart5” and, on this sheet, build two pie charts based on the two existing data tables. The top pie chart should show “Expenditures by Function – All Funds”. Use the amounts from the “TOTAL EXPENDITURES” detail lines for each function in the All Funds summary table, calculate each function’s percentage of total expenditures, and make each function one slice. Label each slice with the function name and its percentage, and add a legend with the same function names. The bottom pie chart should show the “breakdown of General Support Services”. Use the amounts for all rows that belong to the General Support Services function in the detailed table, calculate each item’s percentage of the total General Support Services amount, and make each item one slice. Again, label each slice with the item name and its percentage, and include a matching legend. Both pie charts should be 3-D and visually well-balanced.

{
  "files": [
    "122_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPricing and Valuation · Trading and Risk ManagementSource task tags: Calculation

Original English instruction · agent OS and apps are not declared.

Assume that in January 2002 you enter a “portfolio” by going long ERCOT peak power and short an equal-scale amount of NYMEX natural gas. Over the full year 2002, does this portfolio make or lose money, and what is its nominal return?

{
  "files": [],
  "text": "Interpreting “equal scale” as a dollar-neutral position, the trade loses money over the full year 2002, with a nominal return of approximately –2.0 dollars per portfolio unit."
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational Management · ReportSource task tags: Cross-sheet/file Retrieval · Calculation · Data Entry / Import

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Complete the content in the summary sheet based on other spreadsheets. Leave blank if no information found.

{
  "files": [
    "124_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Calculation · Summary / Visualization

Original English instruction · agent OS and apps are not declared.

You are given an Excel table (Figure 1.19) showing, for IDA-eligible countries, disbursements, amortization, and interest on public and publicly guaranteed debt in 2021–2023 by creditor type (Multilateral, IMF, Bilateral, Commercial bank and other, Bondholders), as well as 2021–2023 totals. At the bottom of the table, the total PPG long-term debt net transfers for 2022 and 2023 are reported. Using only the data in the table, please: Produce at least one chart that displays disbursements, amortization, and interest for each creditor type in 2021–2023. Write a structured economic analysis of no more than 300 words, summarizing the main trends, key turning points, and possible financing behavior patterns. In your analysis, calculate and cite the increase in total long-term debt net transfers in 2023.

{
  "files": [
    "125_ref_0.xlsx"
  ],
  "text": "A suitable visualization is a set of horizontal stacked bar charts by creditor type for 2021–2023, where each bar decomposes disbursements, amortization and interest.\nApplying the net-transfer formula, cumulative 2021–2023 net inflows amount to about US$60 billion from multilateral creditors, US$23 billion from bilateral lenders and roughly US$7 billion from the IMF. In contrast, “Commercial bank and other” and “Bondholders” together show small net outflows of around US$1 billion each over the three-year period, indicating that overall inflows are driven mainly by official sources. Year by year, bondholders shift from a sizeable positive net transfer in 2021 to persistent negative transfers in 2022 and 2023, while commercial banks record a marked net outflow in 2022, representing an important turning point in the composition of flows. Using the total figures at the bottom of the table, PPG long-term debt net transfers rise from about US$21.84 billion in 2022 to roughly US$27.26 billion in 2023, an increase of 24.8%. Overall, the data suggest a clear improvement in aggregate net inflows in 2023, but this recovery is heavily reliant on official creditors, while private financing remains weak, pointing to lingering vulnerabilities in the funding structure."
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationAsset ManagementSource task tags: Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Convert the 'EOTT TX-NM Facilities' worksheet into a standard table and filter it to display Andrew’s county only.

{
  "files": [
    "126_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Calculation · Summary / Visualization

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Complete the Existing Deals with Huber and Proposed Deals with Huber summaries by cross-referencing the Exposure sheet to capture each contract’s start and end dates, delivery point, and contract price. Calculate the average of 55-day payables at each designated time point, and aggregate these amounts into the summary totals corresponding to their respective time periods.

{
  "files": [
    "127_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Summary / Visualization · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared.

Prepare a stacked area chart titled "Existing and Proposed Deals with Huber", showing monthly time on the X-axis (from December 2001 to December 2006) and daily gas supply volumes (MMBtu/d) on the Y-axis in the Summary sheet. Each colored band should represent a distinct pricing benchmark or contract type (deal), illustrating how volumes vary over time

{
  "files": [
    "128_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Summary / Visualization · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared.

Create a stacked area chart titled “Rolling 55 Day Payables (as of the 25th of the Month)” to visualize rolling payable amounts across multiple natural gas price indices from December 2001 to December 2006 in the Summary sheet. The Y-axis should be dollars and the X-axis monthly time, with each data point representing the rolling 55-day payables total as of the 25th of that month.

{
  "files": [
    "129_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationAccounts Payable and ReceivableSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Update the weekly sheet’s rent table to include four new columns — “Total Weekly,” “Total Biweekly,” “Total Monthly,” and “Grand Total” — and populate each tenant’s amounts by week, every two weeks, and monthly, leaving missing items as blank cells. At the bottom, add a “Sub-total” row that sums each column, a “Number of Periods” row reflecting the annual payment count for each tenant based on their rent frequency, and an “Annual Totals” row that calculates the yearly totals for each column; the “Grand Total” column should only be populated on the “Annual Totals” row and remain blank elsewhere.

{
  "files": [
    "130_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationProcurement and SalesSource task tags: Calculation · Validation / Review

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Update the Beaumont Worksheet to reflect a Base Volume assumption of 17,000 (from 18,000) and roll the change through all affected calculations and summaries. Ensure the threshold/labels (e.g., “>17,000”) and the excess/deficit, buyback, swing, baseload, and invoice totals are consistent with the new base.

{
  "files": [
    "131_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Data Entry / Import · Structuring / Formatting · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Update the 'overview' worksheet with the new data located to the right of the 'to be moved to line:' label on the 'support for reclass of ENA' sheet. Also insert a 'TOTAL EWS' line under the top header.

{
  "files": [
    "132_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPlanning and BudgetingSource task tags: Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Audit the consolidated 2002 plan workbook and correct the formula errors and omissions so the subtotals and roll-ups calculate properly.

{
  "files": [
    "133_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Shift the merged title so it begins in column B rather than A to align with the table, and delete all rows that are not marked "violation."

{
  "files": [
    "134_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Summary / Visualization · Structuring / Formatting

Original English instruction · agent OS and apps are not declared.

Create four grey area charts for the AGA Storage Report – Working Gas year-over-year comparison, showing the weekly difference versus the same week last year (in Bcf). Place them on Prod_Yr-Yr (Producing), East_Yr-Yr (East Consuming), West_Yr-Yr (West), and Total_Yr-Yr (Total), with the x-axis running through the week ending 9/23.

{
  "files": [
    "135_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Summary / Visualization · Structuring / Formatting

Original English instruction · agent OS and apps are not declared.

Build four multi-year line charts for “AGA Storage Report – Working Gas,” covering Producing, Consuming East, Consuming West, and Total, that compare weekly Storage Level (Bcf) across 1994/95, 1995/96, 1996/97, 1997/98, 1998/99, and 1999/00. Use Bcf on the Y-axis and Week Ending (from 5-Nov to 21-Oct) on the X-axis, include a vertical dashed line separating WINTER and SUMMER, and place each chart on sheets named Producing, East, West, and Total.

{
  "files": [
    "136_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Update the NC_PL worksheet so its formatting matches the style used on NC_BS.

{
  "files": [
    "137_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Structuring / Formatting · Summary / Visualization

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Add a new worksheet titled “P&C” and build a Property & Casualty Insurance Highlights page modeled after the other summary tabs. Set the header at the top (American Financial Group; Property & Casualty Insurance Highlights; (In millions)) and lay out Fourth Quarter 2003 vs. 2002 and Twelve Months Ended 2003 vs. 2002 with an “Inc (Dec)” column, consistent with the existing format.

{
  "files": [
    "138_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Financial Modeling · Calculation · Summary / Visualization

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Using the Cleburne Plant Damage Sensitivities, evaluate the financial impact at plant capacities of 240MW, 245MW, 258MW, and 263MW. For Q1–Q4, compute and compare the implied equity value at the contract effective date and subsequent dates, determine damages per Section 11.02(i)(a) under current conditions, assess the impact of modifying the interest rate adjustment on the capacity payment, and the result if the interest rate adjustment is fully removed, then quantify the deltas versus the 263MW baseline to show sensitivity of equity value and damages.

{
  "files": [
    "139_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPlanning and BudgetingSource task tags: Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Compute the Monthly Volume and, under different Prime Rate and Corporate Rate assumptions, calculate the monthly Carrying Cost and resulting P/L, then provide the annual roll-up summary.

{
  "files": [
    "140_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Data Entry / Import · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Identify and populate, for each Trader, the corresponding Book Organization for both Power and Natural Gas, using the short-form names only. The update should reflect the abbreviated book codes aligned to each trader (e.g., NETCO power books and PWR-GAS gas books).

{
  "files": [
    "141_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Under the assumptions of Scenario 1, calculate and populate—for each gas point—the total position (contracts), the 30‑day average standard deviation, and the P&L (US$), then aggregate to a grand total for P&L.

{
  "files": [
    "142_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Apply Scenario 2 to calculate the current positions, the 30-day standard deviation and the P&L (US$) for each gas point and then provide the aggregated total for P&L. This should reflect the Scenario 2 convention.

{
  "files": [
    "143_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Calculation · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Using the daily Crude Oil and Natural Gas prices recorded in the NYMEX Settlement Price Averages table, calculate the required averages and the settlement reference prices, and populate the blank fields at the bottom of the table. Format CRUDE OIL values to three decimal places and Natural Gas values to five decimal places. If that particular statistic doesn’t apply to a given column, just fill in “N/A.”

{
  "files": [
    "144_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Web Search · Data Entry / Import · Structuring / Formatting · Calculation · Summary / Visualization

Original English instruction · agent OS and apps are not declared.

I want to compare the labor market between four southern US states, namely Alabama, Georgia, Louisiana, and Mississippi. I need to compile monthly employment data not seasonally adjusted for these states from January 2024 to June 2024 (including January 2024 and June 2024). Please provide the following:\nUnadjusted Unemployment Rate\nLabor Force Participation Rate\nAll Employees – Total Nonfarm (Thousands of Persons)\nAll Employees – Manufacturing (Thousands of Persons)\nAverage Hourly Earnings of All Employees (Total Private)\nAverage Hourly Earnings of All Employees (Manufacturing)\n\nPlease output an xlsx file with Sheet1 as the RawData sheet, with the following column names in order:\nStates, Statistical Month, Unadjusted Unemployment Rate(%), Labor Force Participation Rate(%), All Employees – Total Nonfarm(in thousands), All Employees – Manufacturing(in thousands), Average Hourly Earnings of All Employees (Total Private), Average Hourly Earnings of All Employees (Manufacturing)\nFor Statistical Month, use the format yyyy-mm, like 2024-06.\n\nDon't ask me any questions, just output the results according to the columns without omitting cells arbitrarily.\n\nBased on the RawData sheet, complete the following 4 subtasks:\n\nSubtask 1 - Sheet2_State_Ranking: Rank states by overall labor market health (6-month average).\nColumns: Rank, State, Avg Unemployment Rate (%), Avg Participation Rate (%), Avg Total Employment (thousands), Avg Manufacturing Employment (thousands), Avg Wage - Private ($), Avg Wage - Manufacturing ($), Labor Market Health Score\nHealth Score = (100 - Avg Unemployment) + Avg Participation + (Avg Wage Private - 25). All percentages 2 decimals, employment figures 1 decimal, wages 2 decimals, health score 2 decimals. Sort by health score descending.\n\nSubtask 2 - Sheet3_Monthly_Trends: Calculate month-over-month changes for each state (Feb-Jun, 5 months per state = 20 rows).\nColumns: State, Month, Unemployment Change (pp), Participation Change (pp), Total Employment Growth (%), Manufacturing Employment Growth (%), Wage Growth - Private (%)\nUse formula (Current-Previous)/Previous×100 for growth rates. All values 2 decimals. Sort by State then Month.\n\nSubtask 3 - Sheet4_Employment_Structure: Employment composition analysis for each state (June 2024 snapshot).\nColumns: State, Total Nonfarm Employment (thousands), Manufacturing Employment (thousands), Non-Manufacturing Employment (thousands), Manufacturing Share (%), Total Employment 6M Growth (%), Manufacturing 6M Growth (%)\nNon-Manufacturing = Total - Manufacturing. Share = Manufacturing/Total×100. 6M Growth = (Jun-Jan)/Jan×100. Employment 1 decimal, percentages 2 decimals.\n\nSubtask 4 - Sheet5_Wage_Analysis: Wage premium and volatility analysis by state (6-month statistics).\nColumns: State, Avg Private Wage ($), Avg Manufacturing Wage ($), Wage Premium ($), Wage Premium (%), Wage Volatility - Private ($), Wage Range - Private ($), Premium Category\nPremium = Mfg Wage - Private Wage. Premium % = Premium/Private×100. Volatility = STDEV. Range = MAX - MIN. Category: \"High Premium\" if Premium % > 5%, \"Medium Premium\" if 0-5%, \"Low Premium\" if 0 to -5%, \"Negative Premium\" if < -5%. All wages 2 decimals."

{
  "files": [
    "145_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Web Search · Data Entry / Import · Structuring / Formatting · Calculation · Validation / Review

Original English instruction · agent OS and apps are not declared.

Please compare the growth trajectory of the leading global e-commerce platforms by compiling their quarterly performance for 2024—Amazon, eBay, Shopee (Sea), Alibaba. Use figures from official financial reports for publicly traded companies; for privately held firms, cite reputable news sources and mark relevant numbers with an asterisk (*) to denote that they are estimates. For information you can't find online, just fill in with \"–\".\n\n[Data Source Note] Please use official financial reports for FY2024 (versions published before February 28, 2025). For companies that report in non-USD currencies (e.g., Alibaba), do not recalculate FX yourself—always use the USD figures exactly as stated in the company's official quarterly earnings release, which already apply a fixed exchange rate (the Federal Reserve H.10 rate on or near the quarter-end date). For Alibaba, use the \"Net income\" line (not \"Net income attributable to ordinary shareholders\") as the net profit metric.\n\nPlease output an xlsx file with Sheet1 as the RawData sheet, with the columns (in this order): Company, Quarter, GMV (USD bn), Revenue (USD bn), Net Profit (USD bn)\n\nFor Quarters, label it simply Q1, Q2, etc.\n\nDon't ask me any questions, just output the results according to the column without omitting cells arbitrarily.\n\nBased on the RawData sheet, complete the following 4 subtasks:\n\nSubtask 1 - Sheet2_Financial_Indicators: Calculate key financial performance metrics for each company-quarter.\nColumns: Company, Quarter, Revenue (USD bn), Net Profit (USD bn), Net Profit Margin (%), Monetization Rate (%), Profit per Revenue Dollar ($), Profitability Status\nNet Profit Margin = (Net Profit / Revenue) x 100 (2 decimals). Monetization Rate = (Revenue / GMV) x 100 (2 decimals, \"N/A\" if no GMV). Profit per Revenue Dollar = Net Profit / Revenue (4 decimals). Profitability Status: \"Highly Profitable\" if Net Profit >= $5bn, \"Profitable\" if >= $0, \"Near Breakeven\" if >= -$0.1bn, \"Loss Making\" if < -$0.1bn, \"Unknown\" if missing.\n\nSubtask 2 - Sheet3_Comparison_Matrix: Restructure data into company comparison matrix (pivot structure).\nColumns: Metric, Quarter, Amazon, eBay, Shopee (Sea), Alibaba\nCreate 3 metric blocks (12 rows total): Revenue (USD bn) for Q1-Q4, Net Profit (USD bn) for Q1-Q4, Net Profit Margin (%) for Q1-Q4. All values: 2 decimals, \"N/A\" if unavailable.\n\nSubtask 3 - Sheet4_Growth_Classification: Annual growth analysis and classification by company.\nColumns: Company, Total Revenue (USD bn), Total Net Profit (USD bn), Avg Quarterly Revenue (USD bn), Avg Quarterly Profit (USD bn), Q1-Q4 Revenue Growth (%), Q1-Q4 Profit Change (%), Revenue Growth Category, Profit Trend, Avg Business Model Efficiency\nQ1-Q4 Revenue Growth = (Q4 Revenue - Q1 Revenue) / Q1 Revenue x 100 (2 decimals). Profit Change: output as a percentage string with two decimals (e.g., \"92.31%\"), or \"Turnaround\" if it went from loss to profit. Revenue Growth Category: \"High Growth\" if >=30%, \"Moderate Growth\" if >=10%, \"Stable\" if >=0%, \"Declining\" if <0%. Profit Trend: \"Consistently Improving\" if all quarters increase, \"Declining\" if all decrease, \"Overall Improving\" if Q4>Q1, else \"Volatile\". Business Model Efficiency = Avg Revenue / Avg GMV (4 decimals, \"N/A\" if no GMV).\n\nSubtask 4 - Sheet5_Data_Validation: Validate data quality and identify anomalies.\nColumns: Company, Quarter, Has GMV, Has All Financial Data, Data Completeness (%), Validation Status, Quality Score, Quality Grade\nHas GMV: \"Yes\"/\"No\". Has All Financial Data: \"Yes\" if Revenue and Net Profit exist. Data Completeness = (sum of available fields / 3) x 100 (2 decimals, counting GMV, Revenue, Net Profit). Quality Score: starts from Completeness, -10 if no GMV, -5 per issue (2 decimals). Quality Grade: \"Excellent\" if >=90, \"Good\" if >=70, \"Fair\" if >=50, \"Poor\" if <50.

{
  "files": [
    "146_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk Management · Investment: Credit · Pricing and ValuationSource task tags: Data Entry / Import · Structuring / Formatting · Cross-sheet/file Retrieval · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Fill in the cells highlighted with a blue background, and then remove the background color.

{
  "files": [
    "147_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Structuring / Formatting · Calculation · Financial Modeling

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Switch the timing/rotation model to Method 1 and set the moving average to 120 with a deviation threshold of 35%. After updating the parameters, ensure the model outputs are refreshed to reflect the new settings.

{
  "files": [
    "148_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Calculation · Summary / Visualization

Original English instruction · agent OS and apps are not declared.

Task Background: Global growth is expected to remain steady for the foreseeable future despite escalating geopolitical tensions and heightened uncertainty surrounding global trade policy. Global inflation is expected to moderate, allowing central banks to ease monetary policy to support economic activity. However, the outlook remains subdued for low- and middle-income countries (LMICs), particularly for those with low creditworthiness, constrained fiscal space, and significant political and social unrest. Risks to the macroeconomic outlook for LMICs are tilted to the downside, including the escalation of armed conflicts, further trade fragmentation, persistent global inflation, a weaker global risk appetite, and slower-than-expected growth in major LMICs, especially China. In 2023, LMICs accumulated significant additional debt and faced the corresponding heavy debt service burden. Although global interest rates are declining, debt service costs are expected to moderate gradually. However, LMICs' debt outlook still carries downside risks. The growing prominence of nontraditional creditors, particularly the accumulation of debt owed to China, complicates debt resolution. This is especially critical for small states with underdeveloped domestic financial markets. Furthermore, higher borrowing costs and increased debt service burdens may exacerbate fiscal challenges, especially in LMICs with tightening fiscal policies. Task Objective: Based on the provided spreadsheet data, write a report. The report should include the following: Chart Design and Analysis: For each spreadsheet provided, create at least one chart based on the data. Report Writing: Use the data from the provided spreadsheets to draft the report. Each spreadsheet should be analyzed, and for each one, at least one chart should be generated. Final Output: Save the final report as a PDF document.

{
  "files": [
    "149_ref_0.pdf"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Calculation · Data Entry / Import · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Review and update the Year-End Goal Progress section of the Weekly Commodity Logic Report by tracking the KPIs established at the beginning of the year. Make sure the completion percentages for Potential Revenue-Generating Transactions, $ Financed Through Bank Logic, and Paying Customers Using Production reflect the most current totals.

{
  "files": [
    "150_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPlanning and BudgetingSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Add the necessary rows and columns in the file so EBIT is calculated and shown for the 2001 Forecast, 2001 Proforma, and 2002 Plan.

{
  "files": [
    "151_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Data Entry / Import · Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Update the December entry: set the nominated date to November 22, the Daily Quantity to 24,516, and the Monthly Quantity to 760,000. Add an Estimated Monthly Quantity column, and at the bottom of the table calculate the total for the months already included and the period average daily quantity.

{
  "files": [
    "152_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

On the correlation sheet, add derived columns from the BSCTMP, NBSK, and SBSK price series: a “Dollar difference” showing the spread between SBSK and BSCTMP, and the logarithmic growth rate for each of three price series between two consecutive months; retain the original date column at the end.

{
  "files": [
    "153_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReport · Accounts Payable and ReceivableSource task tags: Data Entry / Import · Calculation · Summary / Visualization

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Complete the missing Interreg co-financing data in the FR finances sheet (Rate, Maximum allocation, Previous payments, Current payment request, and Remaining) for the applicable categories.

{
  "files": [
    "154_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReport · Planning and BudgetingSource task tags: Validation / Review · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Revise the data of 2002 allocation in HR sheet to reflect the figures in HR Detail sheet.

{
  "files": [
    "155_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReport · Planning and BudgetingSource task tags: Data Entry / Import · Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared.

Input the required data into the EA Alloc to Other BUs – Support workbook and complete any missing formulas on the CABC tab (Analysis of I/C Billings).

{
  "files": [
    "156_ref_0.xlsx"
  ],
  "text": ""
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Structuring / Formatting · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Build comprehensive Q2, Q3, and Q4 calculation tables following the same structure and logic as Q1, applying each task’s specific assumptions to compute the Interest Rate Adjustment, Apache Savings, Adjusted Capacity Rate, Months in the Year, Plant Capacity, Yearly Capacity Payments, and Monthly Capacity Payments for each year from 1997 to 2019. Finally, provide the Equity Present Value (XNPV5).

{
  "files": [
    "157_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Validation / Review · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Audit the workbook and correct the formula errors in place so numbers calculate properly.

{
  "files": [
    "158_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPredictive ModelingSource task tags: Structuring / Formatting · Calculation · Financial Modeling · Validation / Review

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Complete the orange-highlighted cells on the Timing Tracking sheet so they consistently point to the Benchmark Returns sheet. On Benchmark Returns, fill in the previously missing columns using the same logic as the existing columns, validate that the annual breakdown is behaving normally, and add formulas to summarize performance across the full sample period with results flowing back to the tracking view. Assume 250 trading days a year for the full sample period.

{
  "files": [
    "159_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationAccounts Payable and Receivable · ReportSource task tags: Data Entry / Import · Structuring / Formatting

Original English instruction · agent OS and apps are not declared.

Transcribe the content from the pdfinto the Excel file.

{
  "files": [
    "160_ref_0.xlsx"
  ],
  "text": ""
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk Management · Other Account ManagementSource task tags: Data Entry / Import · Structuring / Formatting

Original English instruction · agent OS and apps are not declared.

Transcribe the pivot table from the pdf into the Excel file as a single table and add a column as the total of each row.

{
  "files": [
    "161_ref_0.xlsx"
  ],
  "text": ""
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Calculation · Summary / Visualization

Original English instruction · agent OS and apps are not declared.

Based on the provided financial data, generate an asset analysis report (a Word document) for Adidas in 2024, with a focus on the reasons behind the growth of total assets. The report should be divided into two parts: Current Assets and Non-Current Assets, describing their changes and the reasons for their growth.

{
  "files": [
    "162_ref_0.docx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Data Entry / Import · Structuring / Formatting

Original English instruction · agent OS and apps are not declared.

Transcribe the content from the image into the Excel file.

{
  "files": [
    "163_ref_0.xlsx"
  ],
  "text": ""
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Summary / Visualization · Translation

Original English instruction · agent OS and apps are not declared.

Translate the screenshot from the English PDF into Chinese and save it into a single PDF file.

{
  "files": [
    "164_ref_0.pdf"
  ],
  "text": ""
}

{
  "review_flags": [
    "default_source_non_excel_not_in_judge"
  ],
  "review_notes": [
    "Default preprocessing skips source PDF/Markdown/DOCX/images. Sixteen records have such inputs. The top-level pipeline does not forward the opt-in source-non-Excel flag. State the chosen preprocessing policy and retain missing-evidence diagnostics; reference-based judging may still occur, but should not be described as inspection of every input."
  ],
  "review_limit_refs": [
    {
      "issue_id": "FINCH-E02",
      "scope": "workflow",
      "refs": [
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L177-L193",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/preprocessor/preprocessor_main.py#L420-L441",
        "https://github.com/FinWorkBench/Finch/blob/95a8b8d135a528b325be003e54c55f886a22602d/src/prompt_build_pipeline.py#L70-L80"
      ]
    }
  ]
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPlanning and BudgetingSource task tags: Structuring / Formatting · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

In the worksheet “Enron Energy Services – Daily Cash Burn Forecast”, extract all non-empty amount cells for Gas Purchases in the week of 12/10/2001. Write these data into Sheet4 in long format: Column 1: Description Column 2: Date Column 3: Value Ignore blank cells and cells containing “-”. Use conditional formatting to highlight values greater than 600,000 with a red fill.

{
  "files": [
    "165_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationTrading and Risk ManagementSource task tags: Calculation · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Finalize the Position Sensitivities for Gas (in US$) by calculating and populating each gas point’s total position contracts, average stdev and the potential profit or loss of the positions under price movements of ±2 and ±1 standard deviations over the past 30 days.

{
  "files": [
    "166_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationPricing and Valuation · Predictive ModelingSource task tags: Financial Modeling · Calculation

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Based on the assumptions in the table, build out a complete pro forma model for the real estate investment and populate the capital stack and return analysis as shown.

{
  "files": [
    "167_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Insert blank rows between adjacent tables in the workbook to create spacing.

{
  "files": [
    "168_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationAccounts Payable and ReceivableSource task tags: Calculation

Original English instruction · agent OS and apps are not declared.

Using all the amounts listed under “Due from SCs” and “Due to SCs” in the November market-settlement sheets, compute the ISO’s final net exposure over the entire settlement period. You should make use of the following data from both sides: Total Invoiced, Total Collected, Total Paid, and Total Adjustments, and derive the final net receivable or net payable amount based on the complete receivable–payable relationships. Then calculate what percentage this final net amount represents relative to the total invoiced amount on both sides (i.e., the combined Total Invoiced figures). The final answer should report: (1) whether the ISO is net receivable or net payable; and (2) the percentage of the net amount relative to the total invoiced volume, rounded to three decimal places.

{
  "files": [],
  "text": "The ISO is in a net-receivable position, with net receivables totaling $6,704,830.54, which represents approximately 0.493% of the total invoiced amount over the entire settlement period."
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationOperational ManagementSource task tags: Calculation · Data Entry / Import · Cross-sheet/file Retrieval

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

According to the specifications in the Strips sheet, aggregate the daily fundamental data for the PJM Interconnection market and compute the corresponding monthly statistics, suitable for power-trading analysis and review.

{
  "files": [
    "170_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.

Composite financial workflow · artifact evaluationReportSource task tags: Calculation · Data Entry / Import · Structuring / Formatting

Original English instruction · agent OS and apps are not declared. Active constraints are retained below.

Add a sheet Table 1.10, showing Total DEL by departmental group. Total DEL is made up of resource DEL excluding depreciation plus capital DEL. Ignore any discrepancies caused by rounding in the last digit.

{
  "files": [
    "171_ref_0.xlsx"
  ],
  "text": ""
}

Evaluation uses a model-judged rubric; no judge validation was run here.

Reference specifications are preserved; implementation code is linked, not included in the ZIP.