Writing & editing
Words are cheap to generate now, and it shows. Tooling that checks prose against its sources — editing, cleanup, translation.
Shared lane
This lane is carried by both hubs of this survey.
Entries are repositories — libraries, benchmarks, specifications, whatever a ledger admitted as evidence. Some of it is polished; most of it is parts.
Tasks in this category
- Scholarly writing stack 0 of 86 entries
- Write business documents 0 of 86 entries
- Evidence-grounded editing 11 of 86 entries
- De-AI & polish prose 0 of 86 entries
- Translate EN<->ZH 0 of 86 entries
- Redline contracts & documents 0 of 86 entries
Evidence-grounded editing
Campaigns placed on this task
Multiple runs of this task exist; each is its own sealed record — no supersession is implied.
-
2026-07-30__evidence-grounded-editorial-agents
run type (the campaign's own designation) high-recall-map 11 of 86 entries ingested
- Ledger rows
- 55
- Screening rows
- 86
- Repos in ledger
- 11
-
2026-08-01__evidence-grounded-editorial-workflows
run type (the campaign's own designation) high-recall-map Not yet ingested
- Ledger rows
- 56
- Screening rows
- 130
- Repos in ledger
- 1
Entries — 11 of 86 on this hub
-
Writing & editing Live
eugeneyujunhao/mog
Evidence-grounded editing
Editorial draft Official implementation of one of the campaign's anchor works on grounded editing.
Forks 0 Language — Stars 0
Snapshot · retrieved UTC
-
Writing & editing Archived
google-research-datasets/AIS
Evidence-grounded editing
Editorial draft A dataset separating source comprehensibility from attribution; archived upstream, retained with its structured annotations.
Forks 2 Language — Stars 31
Snapshot · retrieved UTC
-
Writing & editing Live
KaijuML/parent
Evidence-grounded editing
Editorial draft Implementation of the PARENT table-aware evaluation metric, separating supported content from coverage.
Forks 5 Language Julia Stars 30
Snapshot · retrieved UTC
-
Writing & editing Live
ParticleMedia/RAGTruth
Evidence-grounded editing
Editorial draft A corpus marking unsupported and contradictory spans at word level across generation tasks; adversarial evidence for grounded editing.
Forks 35 Language Python Stars 260
Snapshot · retrieved UTC
-
Writing & editing Live
princeton-nlp/ALCE
Evidence-grounded editing
Editorial draft Benchmark and code for citation correctness and completeness in generated text.
Forks 51 Language Python Stars 524
Snapshot · retrieved UTC
-
Writing & editing Live
qhjqhj00/WebBrain
Evidence-grounded editing
Editorial draft Official implementation of the WebBrain grounded-generation research.
Forks 5 Language Python Stars 70
Snapshot · retrieved UTC
-
Writing & editing Live
Raldir/FEVEROUS
Evidence-grounded editing
Editorial draft The FEVEROUS fact-verification dataset: mixed sentence and table-cell evidence with support, refute and insufficient labels.
Forks 21 Language PHP Stars 76
Snapshot · retrieved UTC
-
Writing & editing Live
shmsw25/FActScore
Evidence-grounded editing
Editorial draft Atomic-fact precision scoring for long-form output; cited for claim-level editorial evaluation.
Forks 68 Language Python Stars 453
Snapshot · retrieved UTC
-
Writing & editing Live
stanford-oval/storm
Evidence-grounded editing
Editorial draft A research system that collects references, outlines, and writes citation-bearing articles with a polishing stage; six ledger rows cite it.
Forks 2891 Language Python Stars 30876
Snapshot · retrieved UTC
-
Writing & editing Live
TobiasBlask/open-paper-machine
Evidence-grounded editing
Editorial draft A staged drafting, verification and revision pipeline; the ledger grades its implementation evidence narrow and low-confidence, and the entry keeps that framing.
Forks 5 Language Python Stars 16
Snapshot · retrieved UTC
-
Writing & editing Live
zhzihao/WikiGenBench
Evidence-grounded editing
Editorial draft A generation benchmark the ledger records as the same work as another row, with the count inconsistency preserved rather than smoothed.
Forks 2 Language Python Stars 13
Snapshot · retrieved UTC
Tasks with no entry today
5 of 6 tasks in this category carry no entry today. Each is listed with its state: ingested with nothing admitted, on the record and not yet ingested, or planned and not yet run.
- Scholarly writing stack Not yet ingested One campaign on the record; its ledger rows have not been brought into the registry.
- Write business documents Not yet ingested One campaign on the record; its ledger rows have not been brought into the registry.
- De-AI & polish prose Not yet ingested 2 campaigns on the record; their ledger rows have not been brought into the registry.
- Translate EN<->ZH Not yet ingested 2 campaigns on the record; their ledger rows have not been brought into the registry.
- Redline contracts & documents Not yet ingested 2 campaigns on the record; their ledger rows have not been brought into the registry.