No-Worse Context-Aware Decoding: Preventing Neutral Regression in Context-Conditioned Generation

arXiv:2604.16686v1 Announce Type: cross Abstract: Large language models (LLMs) can answer questions and summarize documents when conditioned on external contexts (e.g., retrieved evidence), yet context use remains unreliable: models may overwrite an already-correct output (neutral regression) even when the context is non-informative. We formalize neutral regression as a do-no-harm requirement and quantify it by measuring accuracy drops on baseline-correct items under answer-consistent contexts.