Record · analyzed
Self-Refine: Iterative Refinement with Self-Feedback
One model drafts, critiques and revises its own answer at test time: about 20% better on average across 7 tasks.
1
Listed
Added to the list. How entries were checked when they were listed is in the journal.
2
Collected
| File | Source | SHA-256 |
|---|---|---|
| paper PDF2303.17651.pdf | arxiv.org/pdf/2303.17651 | 81e44592314d80218ad108d3490cd0b84ba5962fefd5c84819987afd77a57087 |
To check a file, download it from its source and compare its SHA-256 with the one above. Collected copies are kept in the editor's RSI system; manifest: design-life/literature/rsi-workflow-improvement/sources.json.
3
Analyzed
Quoted verbatim from design-life/literature/rsi-workflow-improvement/comparison.md (2026-10-07). Parts about applying the paper to the editor's private work are left out.
From “Evidence for improving reusable agent workflows” (2026-10-07)
Inspected edition: Self-Refine, v2, 25 May 2023; PDF pp. 3–5, 9
Mechanism and reported evidence: The same model generates, critiques and revises. Table 1 reports seven tasks with mixed task-specific and preference metrics. GPT-4 math solve rate is 92.9% initially and 93.1% after refinement, a 0.2 percentage-point change. Other tasks show larger gains.
4
Replicated
Not done yet. Not attempted yet.
5
Audited
Not done yet. Not attempted yet.
The editor's view
None recorded yet.