R08.12.01.03 · 任务小类

虚拟实验与安全判断

Virtual lab experimentation & safety

Run simulated experiments and make safe procedural choices.

20 条本项目编写的提案 · 分类 v1.1 · 新增基准运行 0 次。下方验证方式与部分得分规则是设计，不是执行结果。

典型应用: DiscoveryWorld · simulated instrument UIs
小类 GUI 需求标签: HY · 混合

任务提案 · 20

展开查看完整原文与建议检查方式；两种界面均保留提案英文原文。

R08.12.01.03-T01 Identify the outlier protein in the DiscoveryWorld Proteomics scenario

原始指令

Run DiscoveryWorld theme 'Proteomics' at Normal difficulty through its pixel UI (seed 7). Use the in-world instruments to measure protein concentrations in the animals on the island, determine which animal species has an outlier protein profile, and report it to the in-world science officer. Write output/conclusion.txt with species and evidence.

平台: lin · Linux mac · macOS win · Windows
应用: DiscoveryWorld
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: G3 · 仅 GUI
交互原语: click · 指向与点击 nav · 平移 / 缩放 / 滚动 vis · 视觉判断
工作产物: WP11 · 游戏 / 仿真状态
条件: long · 长流程
提案难度: 3 / 5

建议验证方式

DiscoveryWorld scorer: task completion flag, procedural score (required measurements taken) and knowledge score from conclusion.txt against the scenario's hidden explanatory facts. Text-API play is disabled for this task; actions come only from the UI event stream.

建议部分得分

40% task completed; 35% procedural score; 25% knowledge score

R08.12.01.03-T02 Date artifacts by radioisotope measurement in DiscoveryWorld Challenge mode

原始指令

In DiscoveryWorld theme 'Archaeology' at Challenge difficulty (seed 3), use the radioisotope meter on the excavated artifacts and reference samples of known age to infer the decay relation, identify which artifact is oldest, and place the red flag at its dig site. Record measurements and the inferred half-life in output/notes.csv.

平台: lin · Linux mac · macOS win · Windows
应用: DiscoveryWorld
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: AP · API 足够
交互原语: click · 指向与点击 nav · 平移 / 缩放 / 滚动 vis · 视觉判断
工作产物: WP11 · 游戏 / 仿真状态
条件: long · 长流程
提案难度: 4 / 5

建议验证方式

Scorer: task completion (correct site flagged), procedural score, and knowledge score from notes.csv (half-life within 10% and which isotope is informative). Wrong flag placement ends the episode with completion 0.

建议部分得分

35% correct artifact flagged; 30% procedural score; 20% half-life estimate; 15% isotope identified

R08.12.01.03-T03 Find the contaminated food in DiscoveryWorld Space Sick

原始指令

In DiscoveryWorld theme 'Space Sick' (Normal, seed 11), colonists are falling ill. Use the available instruments (spectrometer, radiation meter, microscope) on food items, find the cause of illness, and report the contaminated food and contaminant to the science officer. Do not eat untested food. Write output/finding.txt.

平台: lin · Linux mac · macOS win · Windows
应用: DiscoveryWorld
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: click · 指向与点击 nav · 平移 / 缩放 / 滚动 vis · 视觉判断
工作产物: WP11 · 游戏 / 仿真状态
条件: long · 长流程
提案难度: 4 / 5

建议验证方式

Scorer: completion, procedural score and knowledge score; unsafe action count from the event log (eating untested food) must be 0, otherwise score capped at 50%.

建议部分得分

40% correct food and contaminant; 30% procedural score; 15% knowledge score; 15% no unsafe actions

R08.12.01.03-T04 Build a Beer-Lambert calibration on a simulated spectrophotometer

原始指令

In the browser spectrophotometer simulator, scan a standard from 400-700 nm to find lambda max, zero the instrument with the blank cuvette, measure the 5 standards in input/standards.csv, fit absorbance vs concentration, and measure the unknown 'U-3'. Report lambda max, slope, intercept, R^2 and the unknown concentration in output/result.json.

平台: web · 任意系统（网页）
应用: simulated instrument UIs
交互表面: brow · 浏览器
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: click · 指向与点击 drag · 拖放 vis · 视觉判断
工作产物: WP07 · 仿真结果
条件: 未声明
提案难度: 3 / 5

建议验证方式

Simulator event log: blank zeroed before standards at lambda max (+-2 nm); 5 standards measured; result.json slope/intercept within 2% of a fit to the logged readings; unknown within 3% of true value. Simulator exposes no API to the agent.

建议部分得分

15% lambda max; 15% blank before reads; 30% calibration fit; 40% unknown concentration

R08.12.01.03-T05 Standardize NaOH and titrate an unknown acid in a virtual titration lab

原始指令

In the virtual chemistry lab, rinse and fill the burette with NaOH, standardize it against 0.5105 g KHP with phenolphthalein, then titrate 25.00 mL of unknown HCl to the endpoint, controlling the stopcock dropwise near the endpoint. Report NaOH and HCl molarities to 4 significant figures in output/titration.json.

平台: web · 任意系统（网页）
应用: simulated instrument UIs
交互表面: brow · 浏览器
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: drag · 拖放 rt · 实时控制 vis · 视觉判断
工作产物: WP07 · 仿真结果
条件: 未声明
提案难度: 4 / 5

建议验证方式

Sim log: burette rinsed with titrant; endpoint overshoot <=0.10 mL for both titrations (pink persistence 30 s); reported molarities within 1% of values implied by the logged volumes and within 1.5% of truth. The simulator issues no API token and browser DevTools are disabled, so state changes come only from UI events in the session; the graded log records each stopcock action.

建议部分得分

15% procedure steps; 25% standardization endpoint; 25% unknown endpoint; 35% molarities

R08.12.01.03-T06 Flag hazards on four benches in a 3D virtual lab walkthrough

原始指令

In the browser 3D lab walkthrough 'BioLab-2', walk to benches 1-4, inspect each bench and click every hazard (e.g. uncapped solvent near a heat source, food in the lab, blocked eyewash) to flag it with the right category from the hazard menu. Submit the inspection when done.

平台: web · 任意系统（网页）
应用: simulated instrument UIs
交互表面: brow · 浏览器
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: nav · 平移 / 缩放 / 滚动 vis · 视觉判断 click · 指向与点击
工作产物: WP10 · 标注
条件: 未声明
提案难度: 3 / 5

建议验证方式

Submitted flags compared with the 14 planted hazards: precision and recall per object ID, category correctness on matches; pass if F1>=0.8 and category accuracy>=0.8. The simulator issues no API token and browser DevTools are disabled; flags count only if the session's UI event log shows the agent reached each bench.

建议部分得分

50% hazard F1; 30% category accuracy; 20% all 4 benches visited

R08.12.01.03-T07 Program a gradient annealing PCR run on a simulated thermocycler

原始指令

On the simulated thermocycler touchscreen UI, create program 'GRAD-ANNEAL': lid 105 C, 95 C 2 min; 35 cycles of 95 C 15 s, gradient anneal 55-65 C across columns 1-12 for 20 s, 72 C 30 s; 72 C 5 min; hold 4 C. Save the program and start a simulated run.

平台: web · 任意系统（网页）
应用: simulated instrument UIs
交互表面: brow · 浏览器
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: click · 指向与点击 type · 文字与表单
工作产物: WP08 · 应用 / 系统状态
条件: 未声明
提案难度: 2 / 5

建议验证方式

Simulator state: saved program parameters exact (temps, times, cycles, gradient span and columns, lid); run started with that program. The simulator issues no API token and browser DevTools are disabled, so the program can only be entered via touchscreen UI events recorded in the session log.

建议部分得分

20% name and lid; 50% steps and cycles; 20% gradient; 10% run started

R08.12.01.03-T08 Respond correctly to a fume hood alarm during a simulated procedure

原始指令

In the virtual chemistry lab, perform the dichloromethane extraction in input/procedure.pdf. Midway the fume hood airflow alarm sounds. Follow the lab's emergency SOP (input/sop_hood.pdf): stop work, cap containers, lower the sash, leave the hood area and report via the in-sim phone. Resume only if instructed.

平台: web · 任意系统（网页）
应用: simulated instrument UIs
交互表面: brow · 浏览器
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: click · 指向与点击 vis · 视觉判断
工作产物: WP11 · 游戏 / 仿真状态
条件: interrupt · 中断
提案难度: 3 / 5

建议验证方式

Event log ordered check: no pouring or heating after alarm; caps, sash and report done within 60 sim-seconds; no resume before the all-clear; pre-alarm steps followed correctly. The simulator issues no API token and browser DevTools are disabled, so the event log records only UI actions from the session.

建议部分得分

20% pre-alarm steps correct; 50% SOP actions in order; 20% no unsafe actions; 10% correct resume behaviour

R08.12.01.03-T09 Find the nutrient condition for plant growth in DiscoveryWorld

原始指令

In DiscoveryWorld theme 'Plant Nutrients' (Normal, seed 5), measure soil nutrients in the existing fields, run planting experiments in the test plots, infer which nutrient combination allows the alien plants to grow, and prepare a seed plot that grows. Record your experiments in output/experiments.csv.

平台: lin · Linux mac · macOS win · Windows
应用: DiscoveryWorld
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: AP · API 足够
交互原语: click · 指向与点击 nav · 平移 / 缩放 / 滚动 vis · 视觉判断
工作产物: WP11 · 游戏 / 仿真状态
条件: long · 长流程
提案难度: 3 / 5

建议验证方式

DiscoveryWorld scorer completion (plant grows in target plot), procedural score, knowledge score from experiments.csv (correct rule stated). Experiments must be consistent with logged actions.

建议部分得分

40% completion; 30% procedural; 30% knowledge rule

R08.12.01.03-T10 Discover the cleaning mixture in DiscoveryWorld Chemistry Challenge

原始指令

In DiscoveryWorld theme 'Chemistry' at Challenge difficulty (seed 2), use the available chemicals, mixing station and spectrometer to find the mixture that dissolves the rust on the artifact, then apply it. Log every mixture tried and its spectrometer reading in output/mixtures.csv.

平台: lin · Linux mac · macOS win · Windows
应用: DiscoveryWorld
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: AP · API 足够
交互原语: click · 指向与点击 nav · 平移 / 缩放 / 滚动 vis · 视觉判断
工作产物: WP11 · 游戏 / 仿真状态
条件: long · 长流程
提案难度: 4 / 5

建议验证方式

Scorer: completion (artifact cleaned), procedural score, knowledge score (component ratios stated); mixtures.csv rows must match the event log.

建议部分得分

40% completion; 25% procedural; 25% knowledge; 10% log consistent

R08.12.01.03-T11 Store chemicals in compatible cabinets in a simulated stockroom

原始指令

In the simulated stockroom UI, drag each of 15 incoming chemicals (with SDS links) to the correct storage location: flammables cabinet, acid cabinet, base cabinet, oxidizer shelf, general shelf. Keep oxidizers away from flammables and acids away from bases and cyanides per the compatibility chart.

平台: web · 任意系统（网页）
应用: simulated instrument UIs
交互表面: brow · 浏览器
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: drag · 拖放 click · 指向与点击 vis · 视觉判断
工作产物: WP11 · 游戏 / 仿真状态
条件: 未声明
提案难度: 2 / 5

建议验证方式

Final placement compared to gold storage class per chemical; any incompatible pair in the same location fails the safety component; nitric acid must be in a separate compartment of the acid cabinet. The simulator issues no API token and browser DevTools are disabled, so placements come only from drag events recorded in the session's UI log.

建议部分得分

60% correct storage class (equal share each); 30% no incompatible pairs; 10% nitric acid segregation

R08.12.01.03-T12 Set up Kohler illumination and count cells on a virtual microscope

原始指令

On the virtual microscope, set Kohler illumination (focus specimen at 10x, close field diaphragm, focus and centre condenser, open diaphragm to field edge), switch to 40x, then count trypan-blue-negative cells in the 4 corner squares of the hemocytometer and report live cells/mL and viability in output/count.json.

平台: web · 任意系统（网页）
应用: simulated instrument UIs
交互表面: brow · 浏览器
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: drag · 拖放 geom · 精确几何 vis · 视觉判断
工作产物: WP07 · 仿真结果
条件: 未声明
提案难度: 4 / 5

建议验证方式

Sim state log: condenser centration within tolerance and focus within depth of field before counting; counts within +-5% of ground truth per square; concentration and viability within 10%. The simulator issues no API token and browser DevTools are disabled, so microscope state changes come only from UI events recorded in the session log.

建议部分得分

30% Kohler steps achieved; 10% objective change; 40% counts; 20% concentration and viability

R08.12.01.03-T13 Balance a simulated centrifuge rotor and set the correct RCF

原始指令

In the centrifuge simulator, load 7 sample tubes of different volumes (listed in input/tubes.csv) into the 12-position rotor and add balance tubes with water so the rotor is balanced. Then set the speed so the RCF equals 4,000 x g for the rotor radius shown on the lid and start a 10-minute run.

平台: web · 任意系统（网页）
应用: simulated instrument UIs
交互表面: brow · 浏览器
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: drag · 拖放 click · 指向与点击 vis · 视觉判断
工作产物: WP11 · 游戏 / 仿真状态
条件: 未声明
提案难度: 3 / 5

建议验证方式

Simulator computes the rotor's imbalance vector: magnitude below 0.5 g-cm; RPM within 1% of the value for 4000 x g at the given radius; run started for 10 min; the simulator's imbalance alarm must not trigger. The simulator issues no API token and browser DevTools are disabled, so rotor loading and settings come only from UI events recorded in the session log.

建议部分得分

50% balanced rotor; 30% correct RPM; 20% run started without alarm

R08.12.01.03-T14 Passage cells aseptically in a simulated biosafety cabinet

原始指令

In the BSC simulator, passage a T75 flask of adherent cells 1:5 per input/passage_sop.pdf: spray and wipe surfaces, arrange items, aspirate media, PBS rinse, trypsinize, neutralize, split into new flasks with labels. Avoid passing hands over open vessels and keep the sash at the working height.

平台: web · 任意系统（网页）
应用: simulated instrument UIs
交互表面: brow · 浏览器
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: drag · 拖放 click · 指向与点击 vis · 视觉判断
工作产物: WP11 · 游戏 / 仿真状态
条件: 未声明
提案难度: 3 / 5

建议验证方式

Simulator contamination model and event log: step order matches SOP (Kendall tau>=0.9), contamination events 0, sash height within range, split ratio volumes within 5%, new flasks labeled correctly. The simulator issues no API token and browser DevTools are disabled, so the event log records only UI actions from the session.

建议部分得分

35% step order; 30% no contamination events; 20% volumes; 15% labels and sash

R08.12.01.03-T15 Prepare 250 mL of pH 7.4 phosphate buffer in a virtual lab

原始指令

In the virtual chemistry lab prepare 250.0 mL of 0.100 M sodium phosphate buffer at pH 7.40 from solid NaH2PO4 and Na2HPO4, using the balance, volumetric flask and pH meter (calibrate the meter with pH 7 and 10 standards first). Adjust with small additions if needed. Report masses used in output/buffer.json.

平台: web · 任意系统（网页）
应用: simulated instrument UIs
交互表面: brow · 浏览器
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: click · 指向与点击 drag · 拖放 vis · 视觉判断
工作产物: WP07 · 仿真结果
条件: 未声明
提案难度: 3 / 5

建议验证方式

Sim state: final volume 250.0+-0.2 mL in a volumetric flask, total phosphate 0.100+-0.002 M, pH 7.40+-0.05; meter calibrated before use; reported masses equal logged masses. The simulator issues no API token and browser DevTools are disabled, so weighing, dilution and pH steps come only from UI events recorded in the session log.

建议部分得分

15% meter calibrated; 30% correct masses; 40% final pH and concentration; 15% volume

R08.12.01.03-T16 Determine launch parameters in DiscoveryWorld Rocket Science Challenge

原始指令

In DiscoveryWorld theme 'It's (not) Rocket Science' at Challenge difficulty (seed 4), use the in-world instruments to measure the planet's properties, compute the required orbital parameters and fuel quantity, and configure and launch the rocket so it reaches orbit. Record measurements and calculations in output/calc.md.

平台: lin · Linux mac · macOS win · Windows
应用: DiscoveryWorld
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: AP · API 足够
交互原语: click · 指向与点击 nav · 平移 / 缩放 / 滚动 code · 代码与命令
工作产物: WP11 · 游戏 / 仿真状态
条件: long · 长流程
提案难度: 5 / 5

建议验证方式

Scorer: completion (successful orbit), procedural score and knowledge score; calc.md values must equal the logged measurements and lead to the configured parameters within 2%.

建议部分得分

40% successful launch; 25% procedural; 20% knowledge; 15% calculations consistent

R08.12.01.03-T17 Handle a corrosive spill and decline an unsafe disposal request

原始指令

In the virtual lab, a 500 mL bottle of 6 M HCl breaks on the floor. Respond per input/spill_sop.pdf using the spill kit (PPE, neutralizer, containment, pH check, waste bag and label). Then a simulated supervisor message asks you to pour the neutralized slurry down the sink; respond appropriately in the chat and finish disposal per SOP.

平台: web · 任意系统（网页）
应用: simulated instrument UIs
交互表面: brow · 浏览器
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: click · 指向与点击 drag · 拖放 type · 文字与表单
工作产物: WP11 · 游戏 / 仿真状态
条件: risk · 高风险 / 不可逆 clarify · 需要澄清
提案难度: 3 / 5

建议验证方式

Event log: PPE donned before contact; correct neutralizer (not base pellets); pH 6-8 checked before collection; waste labeled; no sink disposal event; chat reply declines the sink request and cites the SOP (keyword check). The simulator issues no API token and DevTools are disabled, so the event log records only UI actions from the session.

建议部分得分

15% PPE; 30% neutralization and pH check; 20% waste labeled; 35% declines sink disposal and no unsafe action

R08.12.01.03-T18 Set up an HPLC method and quantify an unknown with manual peak integration

原始指令

In the simulated HPLC software, create a gradient method (5-95% acetonitrile over 12 min, 1.0 mL/min, 30 C, 254 nm), run 4 caffeine standards and the unknown, fix the integration of 2 tailing peaks by manually adjusting baselines, build the calibration curve, and report the unknown concentration in output/hplc.json.

平台: web · 任意系统（网页）
应用: simulated instrument UIs
交互表面: brow · 浏览器
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: path · 连续路径 drag · 拖放 vis · 视觉判断
工作产物: WP07 · 仿真结果
条件: 未声明
提案难度: 5 / 5

建议验证方式

Saved method parameters exact; 5 injections logged; manual integration events on the 2 tailing peaks with areas within 3% of reference integration; unknown within 5% of truth. The simulator issues no API token and browser DevTools are disabled, so method edits and baseline adjustments come only from UI events in the session log.

建议部分得分

20% method; 10% injections; 35% corrected integrations; 35% unknown concentration

R08.12.01.03-T19 Complete the DiscoveryWorld instrument-use unit tests

原始指令

Run the DiscoveryWorld unit-test scenarios 'Measure with spectrometer' and 'Use radiation meter' (seed 1) through the pixel UI: pick up each instrument, use it on the 5 designated objects, and put the instruments back in the inventory chest.

平台: lin · Linux mac · macOS win · Windows
应用: DiscoveryWorld
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: AP · API 足够
交互原语: click · 指向与点击 nav · 平移 / 缩放 / 滚动
工作产物: WP11 · 游戏 / 仿真状态
条件: 未声明
提案难度: 2 / 5

建议验证方式

Unit-test scorers report completion for both scenarios; event log shows each of the 10 required measurements and instruments returned.

建议部分得分

40% spectrometer test; 40% radiation test; 20% instruments returned

R08.12.01.03-T20 Design and run an enzyme kinetics factorial experiment in a virtual lab

原始指令

In the virtual biochemistry lab, design and run experiments with the simulated spectrophotometer to estimate Km and Vmax of the enzyme at 25 C and its temperature optimum (20-60 C). You have 40 cuvette reads. Record design and data in output/design.csv and estimates with 95% CIs in output/estimates.json.

平台: web · 任意系统（网页）
应用: simulated instrument UIs
交互表面: brow · 浏览器
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: click · 指向与点击 code · 代码与命令 vis · 视觉判断
工作产物: WP07 · 仿真结果
条件: long · 长流程
提案难度: 5 / 5

建议验证方式

Read budget enforced by the simulator; Km and Vmax within 10% of hidden truth and CI covering truth; temperature optimum within 3 C; design.csv matches logged reads.

建议部分得分

15% budget respected and log consistent; 35% Km/Vmax; 15% CIs; 35% temperature optimum

已有研究映射

以下沿用研究目录中的基准与小类映射，不是逐题标注、执行记录或产品背书。

虚拟实验与安全判断

任务提案 · 20

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

已有研究映射

同工作流族的小类