{"ID":3084807,"CreatedAt":"2026-06-05T06:46:15.197025399Z","UpdatedAt":"2026-06-07T03:05:32.813677833Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2606.05644","arxiv_id":"2606.05644","title":"FIDES: Faithful Inference via Deep Evidence Signals for Retrieval-Memory Conflict in RAG","abstract":"When retrieved evidence contradicts parametric memory, language models frequently ignore context and default to memorized priors -- a failure that undermines the core purpose of retrieval augmentation. Contrastive decoding amplifies the context-conditioned output to suppress parametric bias, but existing methods rest on an implicit assumption that this bias is uniform across tokens. A single global contrastive weight over-penalizes safe tokens while leaving genuinely conflicted ones insufficiently corrected. We identify token-level conflict concentration: retrieval-memory tension is sharply heterogeneous, concentrated on a small fraction of answer-critical decoding steps. This reframes contrastive decoding from how much contrast to apply to where to apply it. We propose FIDES (Faithful Inference via Deep Evidence Signals), a training-free decoder that reads three internal signals probing retrieval-memory conflict at complementary depths -- output surface, hidden representations, and prediction trajectory -- and fuses them to govern intervention strength at each decoding step. Across three benchmarks and six backbones -- four primary 7B/8B models and two scaling backbones up to 70B -- FIDES achieves the best context fidelity in all 18 settings, outperforming the strongest training-free baseline by +3 to +13 points. On the 70B scale, fidelity reaches 92-94% while F1 surges to 62-63%, demonstrating that token-level selectivity unlocks generation capability that coarse contrastive rules suppress.","short_abstract":"When retrieved evidence contradicts parametric memory, language models frequently ignore context and default to memorized priors -- a failure that undermines the core purpose of retrieval augmentation. Contrastive decoding amplifies the context-conditioned output to suppress parametric bias, but existing methods rest o...","url_abs":"https://arxiv.org/abs/2606.05644","url_pdf":"https://arxiv.org/pdf/2606.05644v1","authors":"[\"Zhe Yu\",\"Wenpeng Xing\",\"Tiancheng Zhao\",\"Mohan Li\",\"Changting Lin\",\"Meng Han\"]","published":"2026-06-04T03:16:17Z","proceeding":"cs.AI","tasks":"[\"cs.AI\"]","methods":"[\"Language Model\"]","has_code":false}