Metraipolis › archive › dc_c39ee34ef072
How Much Evidence Should a Coding Agent's Self-Correction Carry? Adaptive Dirichlet Evidence for Self-Distillation — arXiv:2610.08514
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.08514
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T11:19:45.949Z
- SHA-256 of the archived text
68e0d01be5698a34ea94a59d36cf1a053d098b8beafc6608ebd76f20b78ece9b- Stable URL of this page
https://metraipolis.com/documento/dc_c39ee34ef072
How to cite
How Much Evidence Should a Coding Agent's Self-Correction Carry? Adaptive Dirichlet Evidence for Self-Distillation — arXiv:2610.08514 — arXiv (en), https://arxiv.org/abs/2610.08514 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c39ee34ef072 (Metraipolis archive dc_c39ee34ef072, sha256 68e0d01be5698a34ea94a59d36cf1a053d098b8beafc6608ebd76f20b78ece9b)
BibTeX
@misc{dcc39ee34ef072,
title = {How Much Evidence Should a Coding Agent's Self-Correction Carry? Adaptive Dirichlet Evidence for Self-Distillation — arXiv:2610.08514},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.08514},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c39ee34ef072; sha256 68e0d01be5698a34ea94a59d36cf1a053d098b8beafc6608ebd76f20b78ece9b},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.08514. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Execution feedback lets coding agents revise programs and learn from their own corrections. A correction's learning weight should reflect both the transitions supported by its executions and the amount of evidence behind that support. We introduce Effective-Evidence Self-Distillation (EESD), which represents these quantities separately. Normalized execution relevance determines relative transition support and an effective pseudo-count mass; a Dirichlet posterior then produces an uncertainty-penalized weight for KL-anchored correction learning. Under a symmetric prior, changing mass preserves category ordering, and effective mass yields a supervised coefficient bounded by its matched fixed-mass counterpart. Across four model-domain history sweeps, increasing visible observations from one to eight reduces future-outcome NLL by 55.0-59.3%. At eight observations, effective mass achieves lower NLL than fixed mass in all four comparisons. In the primary matched DeepSeek/RunBugRun study, argmax predictions agree on all 3,000 examples, with the largest NLL gain under concentrated relevance. After one correction-learning round, DeepSeek/CodeARC all-tests Pass@1 increases from 15.0% to 20.4%, with a paired 95% source-bootstrap interval of [+2.8, +8.0] percentage points. The twelve-setting downstream evaluation establishes the model-domain scope of this update. These results show how separating evidence support from evidence mass changes probability estimation and correction learning in coding agents. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Artificial Intelligence (cs.AI) ; Software Engineering (cs.SE). Pagina del lavoro: https://arxiv.org/abs/2610.08514 — 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 2026-10-07T11:15:24.617Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).