Jungseob Lee, Seungyoon Lee, Hyeonseok Moon, Seongtae Hong, Suhyune Son, Sugyeong Eo, Minhyuk Kim, Chanjun Park · 2026 · arXiv
DOI: 10.48550/arXiv.2610.29727 · open-access PDF
대규모 언어 모델은 훈련 중에 배운 지식에서 종종 응답합니다.이 문제를 해결하기 위해 대조적 인 디코딩은 모든 토큰에서 전체 및 수정된 인스턴트 아래의 예측을 대조하지만, 또한 통과가 지원하는 올바른 응답을 변경합니다. 우리는 통과에 대한 응답을 확인하는 것이이 거래를 제거한다는 것을 보여줍니다.이 관찰을 바탕으로 우리는 훈련이없는 절차 인 Accept-or-Repair (AoR)를 제안합니다. AoR는 통과가 그것을 지원하는 경우 샘플 된 응답을 유지합니다.
Machine translation (Meta M2M100); the English original follows.
Large language models often answer from knowledge learned during training, even when a provided passage contradicts that knowledge. To address this problem, contrastive decoding contrasts the predictions under the full and a modified prompt at every token, but it also alters correct answers that the passage supports. We show that checking answers against the passage removes this trade-off. Based on this observation, we propose Accept-or-Repair (AoR), a training-free procedure. AoR keeps a sampled answer if the passage supports it. Otherwise, it tries up to two replacements with contrastive dec
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.29727, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange