On GitHub, NousResearch/autoreason has picked up 609 stars, TeX. Autoresearch for subjective domains.
Snapshot summary built from the project's own GitHub metadata — there's no written TopGit review yet. The page will update automatically when a full review is published.
WHY NO REVIEW YET
TopGit writes full reviews for the most-starred, most-requested repositories. This page is a snapshot until then — see the READ ME tab for the original README in full.
Autoreason: Self-Refinement That Knows When to Stop
SHL0MS | HERMES AGENT
Paper (PDF) · Human Eval Materials
Iterative self-refinement fails for three structural reasons: prompt bias (models hallucinate flaws when asked to critique), scope creep (outputs expand unchecked each pass), and lack of restraint (models never say "no changes needed"). Autoreason fixes all three.
Each iteration produces three competing versions — the unchanged incumbent (A), an adversarial revision (B), and a synthesis (AB) — judged by fresh agents with no shared context via blind Borda count. "Do nothing" is always a first-class option.
Key Results
Finding
Detail
42/42 perfect sweep
Haiku 3.5 + autoreason scored perfect Borda across 3 tasks; all baselines degraded below single-pass
77% vs 73%
Sonnet 4.6 on 150 CodeContests problems (private-test), autoreason vs single-pass
40% vs 31%
Haiku 3.5 autoreason vs best-of-6 sampling at matched compute (150 problems)
Haiku 4.5: transition point
At 60% private accuracy, autoreason's held-out gains vanish — the generation-evaluation gap has closed
Blinded materials for human raters are in human_eval/. 5 tasks × 3 methods (autoreason, critique-and-revise, single-pass), randomized 4-character codes. See human_eval/README.md for the rubric and instructions.
Citation
@article{shl0ms2026autoreason,
title={Autoreason: Self-Refinement That Knows When to Stop},
author={SHL0MS and Hermes Agent},
year={2026},
url={https://github.com/NousResearch/autoreason}
}
Does NousResearch/autoreason have a project website?
No homepage URL was recorded for NousResearch/autoreason in TopGit's last sync. The README tab above frequently contains screenshots and demo links, or check the repository description on GitHub.
How many stars does NousResearch/autoreason have?
NousResearch/autoreason has 609 GitHub stars — refresh the page for the live number, or check github.com/NousResearch/autoreason. TopGit mirrors GitHub's count but does not claim minute-by-minute accuracy.
Is NousResearch/autoreason open source?
TopGit's metadata for NousResearch/autoreason does not record a license. Most public repositories on GitHub ARE open source, but the exact terms vary — verify by opening the LICENSE file directly.
What is NousResearch/autoreason?
NousResearch/autoreason (NousResearch/autoreason) is a TeX project on GitHub. From the project's own README: Autoresearch for subjective domains.
Where do I read more about NousResearch/autoreason?
This TopGit page is a snapshot — the READ ME tab shows the project's own README content (links stripped, images preserved). The GitHub repository at github.com/NousResearch/autoreason is the definitive source.
Read full README in the tab above.
Want a second opinion on autoreason?
Ask an AI that can read this page — one click and you get its take on autoreason.