Metraipolis › archive › dc_4cfb582569ba
When Should LLMs Trust Their Own Revisions? A Risk-Aware Study of Intrinsic Self-Correction — arXiv:2609.35832
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.35832
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T16:24:48.335Z
- SHA-256 of the archived text
9f03935294f7831516193aa345241c885698e2d8aec11299d06c7e1b6ff2baef- Stable URL of this page
https://metraipolis.com/documento/dc_4cfb582569ba
How to cite
When Should LLMs Trust Their Own Revisions? A Risk-Aware Study of Intrinsic Self-Correction — arXiv:2609.35832 — arXiv (en), https://arxiv.org/abs/2609.35832 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4cfb582569ba (Metraipolis archive dc_4cfb582569ba, sha256 9f03935294f7831516193aa345241c885698e2d8aec11299d06c7e1b6ff2baef)
BibTeX
@misc{dc4cfb582569ba,
title = {When Should LLMs Trust Their Own Revisions? A Risk-Aware Study of Intrinsic Self-Correction — arXiv:2609.35832},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.35832},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4cfb582569ba; sha256 9f03935294f7831516193aa345241c885698e2d8aec11299d06c7e1b6ff2baef},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.35832. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Intrinsic self-correction asks a language model to revise its own answer without receiving new external evidence. A second pass can recover mistakes, but it can also overturn answers that were already correct. We study this trade-off across 29 open-weight LLMs on BoolQ, GSM8K, and Corr2Cause by tracking correctness transitions between initial and revised answers. Aggregate accuracy can conceal substantially different revision behavior: for example, Llama-3.1-8B improves by 25.5 percentage points on GSM8K, while refinement changes 19.1% of initially correct answers into wrong ones. A controlled BoolQ study further shows that refinement prompts shift the balance between recovery and harm. We then compare three runtime choices: keeping the initial answer, always accepting the revision, and selectively invoking revision using signals available after the initial response. The comparison identifies settings where learned gating is useful and others where a simpler unconditional policy performs better. These results suggest treating intrinsic self-correction as a revision policy rather than as a uniformly beneficial second pass, and evaluating it through both the corrections it recovers and the errors it introduces. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.AI). Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.35832 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-07T15:54:53.719Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).