@inproceedings{yu-etal-2026-fides,
title = "{FIDES}: Faithful Inference via Deep Evidence Signals for Retrieval-Memory Conflict in {RAG}",
author = "Yu, Zhe and
Xing, Wenpeng and
Zhao, Tiancheng and
Li, Mohan and
Lin, Changting and
Han, Meng",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1772/",
pages = "40386--40405",
ISBN = "979-8-89176-459-0",
abstract = "When retrieved evidence contradicts parametric memory, language models frequently ignore the context and revert to memorized priors{---}a failure that undermines the purpose of retrieval augmentation. Contrastive decoding amplifies context-conditioned output to suppress parametric bias, but fixed coefficients cannot track changing disagreement, while output-only adaptive coefficients observe only the final prediction distributions. We propose FIDES (Faithful Inference via Deep Evidence Signals), a training-free decoder that combines signals from output distributions, hidden representations, and prediction trajectories to recompute contrastive intervention strength at every step. A matched, formula-verified audit on 128 NQ-Swap pairs shows that conflict raises Opposition by 0.0969 and the resulting coefficient {\ensuremath{\alpha}} by 0.0727 at the same answer-prediction position (both p {\ensuremath{<}} 10{\ensuremath{^-}}{\ensuremath{^4}}). The audit also finds only a small Shift effect and saturated clipped Noise, clarifying the limits of the mechanism interpretation. 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. At 70B, fidelity reaches 92{--}94{\%} while F1 reaches 62{--}63{\%}, showing that multi-depth stepwise contrast can preserve generation utility while improving evidence following."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yu-etal-2026-fides">
<titleInfo>
<title>FIDES: Faithful Inference via Deep Evidence Signals for Retrieval-Memory Conflict in RAG</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zhe</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wenpeng</namePart>
<namePart type="family">Xing</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tiancheng</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohan</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Changting</namePart>
<namePart type="family">Lin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Meng</namePart>
<namePart type="family">Han</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>When retrieved evidence contradicts parametric memory, language models frequently ignore the context and revert to memorized priors—a failure that undermines the purpose of retrieval augmentation. Contrastive decoding amplifies context-conditioned output to suppress parametric bias, but fixed coefficients cannot track changing disagreement, while output-only adaptive coefficients observe only the final prediction distributions. We propose FIDES (Faithful Inference via Deep Evidence Signals), a training-free decoder that combines signals from output distributions, hidden representations, and prediction trajectories to recompute contrastive intervention strength at every step. A matched, formula-verified audit on 128 NQ-Swap pairs shows that conflict raises Opposition by 0.0969 and the resulting coefficient \ensuremathα by 0.0727 at the same answer-prediction position (both p \ensuremath< 10\ensuremath⁻\ensuremath⁴). The audit also finds only a small Shift effect and saturated clipped Noise, clarifying the limits of the mechanism interpretation. 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. At 70B, fidelity reaches 92–94% while F1 reaches 62–63%, showing that multi-depth stepwise contrast can preserve generation utility while improving evidence following.</abstract>
<identifier type="citekey">yu-etal-2026-fides</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1772/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>40386</start>
<end>40405</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T FIDES: Faithful Inference via Deep Evidence Signals for Retrieval-Memory Conflict in RAG
%A Yu, Zhe
%A Xing, Wenpeng
%A Zhao, Tiancheng
%A Li, Mohan
%A Lin, Changting
%A Han, Meng
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F yu-etal-2026-fides
%X When retrieved evidence contradicts parametric memory, language models frequently ignore the context and revert to memorized priors—a failure that undermines the purpose of retrieval augmentation. Contrastive decoding amplifies context-conditioned output to suppress parametric bias, but fixed coefficients cannot track changing disagreement, while output-only adaptive coefficients observe only the final prediction distributions. We propose FIDES (Faithful Inference via Deep Evidence Signals), a training-free decoder that combines signals from output distributions, hidden representations, and prediction trajectories to recompute contrastive intervention strength at every step. A matched, formula-verified audit on 128 NQ-Swap pairs shows that conflict raises Opposition by 0.0969 and the resulting coefficient \ensuremathα by 0.0727 at the same answer-prediction position (both p \ensuremath< 10\ensuremath⁻\ensuremath⁴). The audit also finds only a small Shift effect and saturated clipped Noise, clarifying the limits of the mechanism interpretation. 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. At 70B, fidelity reaches 92–94% while F1 reaches 62–63%, showing that multi-depth stepwise contrast can preserve generation utility while improving evidence following.
%U https://aclanthology.org/2026.emnlp-main.1772/
%P 40386-40405
Markdown (Informal)
[FIDES: Faithful Inference via Deep Evidence Signals for Retrieval-Memory Conflict in RAG](https://aclanthology.org/2026.emnlp-main.1772/) (Yu et al., EMNLP 2026)
ACL