Metraipolis › archive › dc_020b720e2541
Better Answers, Weaker Evidence Control: Evidence Bypass in On-Policy Distillation — arXiv:2611.01967
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.01967
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-03T10:36:17.851Z
- SHA-256 of the archived text
113a55533504698416960b2e9026b5ac115abb3c8c37dfe263e85b112477d26a- Stable URL of this page
https://metraipolis.com/documento/dc_020b720e2541
How to cite
Better Answers, Weaker Evidence Control: Evidence Bypass in On-Policy Distillation — arXiv:2611.01967 — arXiv (en), https://arxiv.org/abs/2611.01967 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_020b720e2541 (Metraipolis archive dc_020b720e2541, sha256 113a55533504698416960b2e9026b5ac115abb3c8c37dfe263e85b112477d26a)
BibTeX
@misc{dc020b720e2541,
title = {Better Answers, Weaker Evidence Control: Evidence Bypass in On-Policy Distillation — arXiv:2611.01967},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.01967},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_020b720e2541; sha256 113a55533504698416960b2e9026b5ac115abb3c8c37dfe263e85b112477d26a},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.01967. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
On-policy distillation (OPD) can improve answer accuracy while weakening responsiveness to supporting evidence. We investigate this separation in scientific-document question answering through matched document interventions that change the supported answer while preserving the question and unrelated context. We identify evidence bypass: a student retains an originally correct answer after the document changes to support a different one. On QASPER with a Qwen3-4B student, standard OPD raises original-document accuracy by 3.4 percentage points, lowers counterfactual accuracy by 5.3 points, and increases bypass by 15.7 points; the same directional pattern appears on PeerQA. Controlled perturbations localize the shift to answer-critical evidence, and bypass grows with the student's pre-distillation closed-book preference for the original answer. A factorized study of teacher targets, privileged information, divergence, and supervision density finds that explicitly supplying the teacher with annotated answer-critical spans yields the largest joint improvement in evidence response and claim support. These findings motivate Evidence-Paired OPD (EP-OPD), which combines on-policy supervision on matched documents supporting different answers to the same question with an answer-level margin that favors each document's supported answer over its paired alternative. Across both datasets and two student sizes, EP-OPD recovers 54-66% of standard OPD's counterfactual-accuracy loss while preserving 80-85% of its original-document accuracy gain. Our results establish evidence responsiveness as a distinct training and evaluation target alongside answer accuracy. Our code is publicly available at: this https URL . Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2611.01967 — 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 2026-11-03T10:29:48.826Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).