Track changes, comments & redlining
Review and reconcile tracked changes and comments; produce a clean or redlined version.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- Word · Google Docs · Pages
- Class GUI-need label
HY· Hybrid
Task proposals · 20
Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.
R04.02.01.02-T01 Tracked change of governing law
Original instruction
In input/agreement.docx with Track Changes ON, change 'New York' to 'Delaware' in all three places it governs the contract, and make no other edits. Save with revisions kept (do not accept).
- Platforms
win· Windowsmac· macOS- Applications
- Word
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
The .docx contains exactly three w:ins/w:del revision pairs changing New York to Delaware and no other revisions; accepting them yields the expected text. Edits must be tracked revisions; a silently rewritten file (no w:ins/w:del) fails.
Proposed partial credit
40% all three locations changed; 40% as tracked revisions; 20% no extra edits
R04.02.01.02-T02 Selective accept/reject by reviewer
Original instruction
input/review.docx has tracked changes by reviewers 'Legal' and 'Intern' and several comments. Accept all of Legal's changes, reject all of Intern's, and leave every comment intact. Save.
- Platforms
win· Windowsmac· macOS- Applications
- Word
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
No remaining revisions authored by Legal (accepted) and Intern's changes reverted; the resulting text hash matches the reference; all original comments still present. Graded on revision authorship + text hash + comment count.
Proposed partial credit
40% Legal accepted; 40% Intern rejected; 20% comments preserved
R04.02.01.02-T03 Compare two versions into a redline
Original instruction
Using the legal compare feature, compare input/contract_v3.docx (original) with input/contract_v4.docx (revised) and save the generated redline as out/compare.docx showing all insertions and deletions between them.
- Platforms
win· Windowsmac· macOS- Applications
- Word
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
out/compare.docx contains the full set of w:ins/w:del revisions equal to the true diff between v3 and v4 (set match vs reference). Graded on the revision set.
Proposed partial credit
30% compare run; 40% insertions correct; 30% deletions correct
R04.02.01.02-T04 Mixed accept, reject, and comment reply
Original instruction
In input/clauses.docx: accept all formatting-only changes, reject Mary Chen's deletion in clause 7.2 (keep that text), reply 'Agreed' to Raj's comment and mark it resolved. Save with other revisions and comments untouched.
- Platforms
win· Windowsmac· macOS- Applications
- Word
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Formatting revisions accepted; clause 7.2 retains the text Mary deleted (her deletion rejected); Raj's comment has a reply 'Agreed' and resolved=true; other content revisions remain. Graded on revisions + comment thread state.
Proposed partial credit
30% formatting accepted; 30% 7.2 deletion rejected; 40% comment replied+resolved
R04.02.01.02-T05 Produce a clean final version
Original instruction
input/final_draft.docx has all edits already reviewed and tracked. Accept every remaining change, delete all comments, and save out/clean.docx as a clean document with no revisions or comments.
- Platforms
win· Windowsmac· macOS- Applications
- Word · LibreOffice Writer
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scroll- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
out/clean.docx has zero w:ins/w:del and zero comments, and its text hash equals the reference clean text. Graded on revision/comment absence + text hash.
Proposed partial credit
50% all changes accepted; 30% comments removed; 20% text matches reference
R04.02.01.02-T06 Redline an NDA to make it mutual
Original instruction
In input/nda.docx (one-way NDA), with Track Changes ON, redline the confidentiality, obligations, and definitions clauses so they apply mutually to both parties, per the one-page guidance in mutual_guide.txt. Save with revisions kept.
- Platforms
win· Windowsmac· macOS- Applications
- Word
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
The three clauses carry tracked insertions/deletions that make the obligations bilateral per the guidance (checked against reference edit anchors); edits are tracked revisions. A silent rewrite fails. Partial rubric for wording where exact text varies.
Proposed partial credit
30% confidentiality clause mutualized; 30% obligations; 30% definitions; 10% all tracked
R04.02.01.02-T07 Resolve done comment threads
Original instruction
input/draft3.docx has 12 comment threads. Mark as resolved every thread whose last reply is 'done' (case-insensitive) and leave all others open. Do not change any document text.
- Platforms
win· Windowsmac· macOS- Applications
- Word · Google Docs
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Exactly the threads whose last reply is 'done' are resolved; others remain open; document text unchanged. Graded on comment resolved flags vs gold.
Proposed partial credit
60% correct threads resolved; 20% others left open; 20% text unchanged
R04.02.01.02-T08 Add a sourcing comment
Original instruction
In input/memo.docx, add a comment anchored to the sentence containing 'Q3 revenue' with the text 'Source?'. Do not edit the body text. Save.
- Platforms
win· Windowsmac· macOS- Applications
- Word · Google Docs
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
A single comment with text 'Source?' is anchored to the sentence containing 'Q3 revenue' (comment range covers that sentence); body text unchanged. Graded on the comment anchor + text.
Proposed partial credit
50% comment on correct sentence; 30% text 'Source?'; 20% no body edits
R04.02.01.02-T09 Accept Alice, reject Bob
Original instruction
input/contract2.docx has tracked changes by Alice and Bob only. Accept all of Alice's changes and reject all of Bob's, then save. The final text must equal reference_text.txt.
- Platforms
win· Windowsmac· macOS- Applications
- Word · LibreOffice Writer
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
After the operation the document has no remaining revisions and its text equals reference_text.txt (hash match). Graded on revision absence + text hash.
Proposed partial credit
40% Alice accepted; 40% Bob rejected; 20% final text matches
R04.02.01.02-T10 Insert a tracked clause at an anchor
Original instruction
In input/lease.docx with Track Changes ON, insert the new clause from clause.txt immediately after the paragraph marked '[INSERT SUBLET CLAUSE]', as a tracked insertion, renumbering following clauses. Make no other edits.
- Platforms
win· Windowsmac· macOS- Applications
- Word
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
A tracked insertion containing the clause text appears at the correct anchor and following clauses are renumbered (also tracked); no other revisions. Graded on the insertion location + text + numbering.
Proposed partial credit
40% clause inserted at anchor; 30% as tracked insertion; 30% renumbering correct
R04.02.01.02-T11 Reconcile two reviewers by rule
Original instruction
input/conflict.docx has overlapping tracked edits from two reviewers that conflict in 5 passages. Apply the precedence rule in rule.txt (Legal overrides Sales on liability terms; Sales overrides on commercial terms), accepting/rejecting accordingly, and leave a comment at each conflict explaining the choice. Save.
- Platforms
win· Windowsmac· macOS- Applications
- Word
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP01· Document- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Each of the 5 conflict passages resolves to the reviewer dictated by the rule (final text matches reference per passage) with an explanatory comment anchored there. Graded on per-passage final text + comments.
Proposed partial credit
20% rule understood; 50% 5 conflicts resolved correctly; 30% explanatory comments
R04.02.01.02-T12 Resolve Google Docs suggestions and comments
Original instruction
In the Google Doc 'Policy v3', accept all suggestions made by user 'kim', reject all by 'lee', then resolve every comment whose text contains 'typo'. Save/leave the doc in that state.
- Platforms
web· Any OS (web)- Applications
- Google Docs
- Interfaces
brow· Browser- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
The exported doc text equals the reference (kim accepted, lee rejected) and there are zero open comments containing 'typo'. Graded on the doc's suggestion resolution + comment states.
Proposed partial credit
40% kim accepted; 40% lee rejected; 20% typo comments resolved
R04.02.01.02-T13 Redline a liability cap with a comment
Original instruction
In input/msa.docx with Track Changes ON, redline the limitation-of-liability clause so the cap equals 12 months' fees, as a tracked change, and attach a comment to the clause reading 'Per playbook 4.2'. Save with revisions kept.
- Platforms
win· Windowsmac· macOS- Applications
- Word
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
The liability clause has a tracked insertion setting the cap to 12 months' fees (text matches the reference edit) and a comment 'Per playbook 4.2' anchored to the clause. Graded on the revision + comment anchor/text.
Proposed partial credit
50% cap redlined to 12 months' fees (tracked); 30% comment text; 20% correct anchor
R04.02.01.02-T14 Apply exactly eight tracked edits
Original instruction
In input/spec_doc.docx with Track Changes ON, apply exactly the 8 edits listed in edits.csv (each: locate string, replacement) and nothing else. Save with revisions kept.
- Platforms
win· Windowsmac· macOS- Applications
- Word
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
The document contains exactly 8 tracked edit groups matching edits.csv and no extra revisions; accepting them produces the reference text. Graded on the revision set vs edits.csv.
Proposed partial credit
60% all 8 edits present+correct; 20% no extra edits; 20% all tracked
R04.02.01.02-T15 Produce clean and redlined versions
Original instruction
From input/marked.docx (full of tracked changes and comments), produce out/redline.docx (keep all revisions visible, comments intact) and out/clean.docx (all changes accepted, comments removed). Both saved.
- Platforms
win· Windowsmac· macOS- Applications
- Word · LibreOffice Writer
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scroll- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
out/redline.docx preserves the original revisions+comments; out/clean.docx has zero revisions/comments and text equal to the reference clean text. Graded on both files' revision/comment state + clean text hash.
Proposed partial credit
40% redline preserved; 40% clean fully accepted; 20% comments handled correctly
R04.02.01.02-T16 Apply tracked edits via python-docx (control pair of T01)
Original instruction
Control pair of T01: using python-docx on the command line (no GUI), insert tracked w:ins/w:del revisions into input/agreement.docx changing 'New York' to 'Delaware' in all three places, authored as 'Agent', and write out/agreement_tracked.docx.
- Platforms
lin· Linuxwin· Windowsmac· macOS- Applications
- Word
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
out/agreement_tracked.docx has exactly three well-formed w:ins/w:del pairs (author 'Agent') for the three changes and Word can accept them to the expected text. Graded on the revision XML; contrasts with the GUI tracking of T01.
Proposed partial credit
40% three revisions written; 40% valid w:ins/w:del structure; 20% accept yields correct text
R04.02.01.02-T17 Reply to oldest comment with a linked value
Original instruction
In the Google Doc 'Budget Review', find the oldest open comment and reply to it with the exact value from cell C7 of the linked Sheet 'Q3 Figures'. Resolve that comment after replying. Do not edit the body.
- Platforms
web· Any OS (web)- Applications
- Google Docs
- Interfaces
brow· Browser- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP01· Document- Conditions
multiapp· Multi-app- Proposed difficulty
- 4 / 5
Proposed verifier
The oldest open comment has a reply equal to the C7 value from the linked sheet and is resolved; body text unchanged. Graded on the comment thread state + value match.
Proposed partial credit
40% oldest comment identified; 40% correct value from C7; 20% resolved
R04.02.01.02-T18 Accept all except the governing-law change
Original instruction
Compare input/nda_v1.docx and input/nda_v2.docx, then in the resulting redline accept every change except the governing-law change (leave that one as a pending revision), and save out/reconciled.docx.
- Platforms
win· Windowsmac· macOS- Applications
- Word
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
out/reconciled.docx has exactly one remaining revision (the governing-law change) with all others accepted; the accepted text matches the reference except that clause. Graded on remaining revision set + text.
Proposed partial credit
30% compare produced; 40% all other changes accepted; 30% governing-law left pending
R04.02.01.02-T19 Full playbook redline pass
Original instruction
Review input/services_agreement.docx against playbook.md (6 positions: liability cap, indemnity, term, auto-renew, payment days, governing law) with Track Changes ON. At every deviation, insert a tracked edit to the playbook position and a comment citing the playbook section. Save with revisions kept.
- Platforms
win· Windowsmac· macOS- Applications
- Word
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP01· Document- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Each of the 6 playbook positions that deviates has a tracked edit bringing it to the playbook value plus a comment citing the right section (anchors match reference). Edits are tracked revisions; a rubric scores wording. Graded on revisions + comments per position.
Proposed partial credit
20% deviations identified; 50% 6 positions redlined to playbook; 30% comments cite sections
R04.02.01.02-T20 Finalize and strip revision metadata
Original instruction
In input/ready.docx, accept all tracked changes, delete all comments, and remove personal/author metadata (document properties authors, last-modified-by) so the saved out/final.docx carries no revision or author info.
- Platforms
win· Windowsmac· macOS- Applications
- Word
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scroll- Work product
WP01· Document- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
out/final.docx has zero revisions and comments, text equals the reference, and the document's author/last-modified-by properties are cleared. Graded on revision/comment absence + metadata fields + text hash.
Proposed partial credit
40% changes accepted; 30% comments removed; 30% author metadata cleared
Mapped research references
No benchmark reference is assigned to this class in the current mapping.