@inproceedings{asad-etal-2026-faithfulness,
title = "Faithfulness Is Not Free: Auditing Offline {KV}-Cache Quantization in Retrieval-Augmented Generation",
author = "Asad, Atta UL and
Bilal, Ahsan and
Ali, Muhammad and
Haseeb, Muhammad and
Hougen, Dean F.",
editor = "Wang, Yimu and
Choi, Yee Man and
Wu, Di and
Ouyang, Siqi and
Azadani, Mozhgan Nasr and
Fung, Yi R. (May)",
booktitle = "Proceedings of the 1st Workshop on Grounding Language Models: Learning Faithfully and Efficiently ({G}round{LM} 2026)",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.groundlm-1.38/",
pages = "472--479",
ISBN = "979-8-89176-478-1",
abstract = "Retrieval-augmented generation (RAG) systems can precompute and store key-value (KV) caches of retrieved documents to avoid re-encoding the same context for every query. Quantizing these offline caches reduces storage cost, but existing work largely evaluates compression through accuracy alone and does not ask whether the generated answers remain faithful to the retrieved evidence. This distinction is important because a model can preserve the correct answer string while weakening its grounding in the provided context. We audit Qwen2.5-7B-Instruct under BF16, INT8, and INT4 offline KV-cache settings on RGB and HotpotQA, measuring both answer accuracy and faithfulness using a hallucination detector, NLI entailment, and an LLM judge. INT8 is nearly lossless across accuracy and faithfulness metrics. In contrast, INT4 reduces accuracy and, more critically, degrades faithfulness even on examples where the answer remains factually correct: over 90{\%} of faithfulness flips are negative. The degradation increases under noisier retrieval and larger retrieval depth. These results show that compressed RAG systems cannot be validated by EM or F1 alone; faithfulness must be audited before low-bit offline KV caches are deployed."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="asad-etal-2026-faithfulness">
<titleInfo>
<title>Faithfulness Is Not Free: Auditing Offline KV-Cache Quantization in Retrieval-Augmented Generation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Atta</namePart>
<namePart type="given">UL</namePart>
<namePart type="family">Asad</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ahsan</namePart>
<namePart type="family">Bilal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Muhammad</namePart>
<namePart type="family">Ali</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Muhammad</namePart>
<namePart type="family">Haseeb</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dean</namePart>
<namePart type="given">F</namePart>
<namePart type="family">Hougen</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 1st Workshop on Grounding Language Models: Learning Faithfully and Efficiently (GroundLM 2026)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yimu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yee</namePart>
<namePart type="given">Man</namePart>
<namePart type="family">Choi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Di</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Siqi</namePart>
<namePart type="family">Ouyang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mozhgan</namePart>
<namePart type="given">Nasr</namePart>
<namePart type="family">Azadani</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yi</namePart>
<namePart type="given">R</namePart>
<namePart type="given">(May)</namePart>
<namePart type="family">Fung</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">979-8-89176-478-1</identifier>
</relatedItem>
<abstract>Retrieval-augmented generation (RAG) systems can precompute and store key-value (KV) caches of retrieved documents to avoid re-encoding the same context for every query. Quantizing these offline caches reduces storage cost, but existing work largely evaluates compression through accuracy alone and does not ask whether the generated answers remain faithful to the retrieved evidence. This distinction is important because a model can preserve the correct answer string while weakening its grounding in the provided context. We audit Qwen2.5-7B-Instruct under BF16, INT8, and INT4 offline KV-cache settings on RGB and HotpotQA, measuring both answer accuracy and faithfulness using a hallucination detector, NLI entailment, and an LLM judge. INT8 is nearly lossless across accuracy and faithfulness metrics. In contrast, INT4 reduces accuracy and, more critically, degrades faithfulness even on examples where the answer remains factually correct: over 90% of faithfulness flips are negative. The degradation increases under noisier retrieval and larger retrieval depth. These results show that compressed RAG systems cannot be validated by EM or F1 alone; faithfulness must be audited before low-bit offline KV caches are deployed.</abstract>
<identifier type="citekey">asad-etal-2026-faithfulness</identifier>
<location>
<url>https://aclanthology.org/2026.groundlm-1.38/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>472</start>
<end>479</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Faithfulness Is Not Free: Auditing Offline KV-Cache Quantization in Retrieval-Augmented Generation
%A Asad, Atta UL
%A Bilal, Ahsan
%A Ali, Muhammad
%A Haseeb, Muhammad
%A Hougen, Dean F.
%Y Wang, Yimu
%Y Choi, Yee Man
%Y Wu, Di
%Y Ouyang, Siqi
%Y Azadani, Mozhgan Nasr
%Y Fung, Yi R. (May)
%S Proceedings of the 1st Workshop on Grounding Language Models: Learning Faithfully and Efficiently (GroundLM 2026)
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-478-1
%F asad-etal-2026-faithfulness
%X Retrieval-augmented generation (RAG) systems can precompute and store key-value (KV) caches of retrieved documents to avoid re-encoding the same context for every query. Quantizing these offline caches reduces storage cost, but existing work largely evaluates compression through accuracy alone and does not ask whether the generated answers remain faithful to the retrieved evidence. This distinction is important because a model can preserve the correct answer string while weakening its grounding in the provided context. We audit Qwen2.5-7B-Instruct under BF16, INT8, and INT4 offline KV-cache settings on RGB and HotpotQA, measuring both answer accuracy and faithfulness using a hallucination detector, NLI entailment, and an LLM judge. INT8 is nearly lossless across accuracy and faithfulness metrics. In contrast, INT4 reduces accuracy and, more critically, degrades faithfulness even on examples where the answer remains factually correct: over 90% of faithfulness flips are negative. The degradation increases under noisier retrieval and larger retrieval depth. These results show that compressed RAG systems cannot be validated by EM or F1 alone; faithfulness must be audited before low-bit offline KV caches are deployed.
%U https://aclanthology.org/2026.groundlm-1.38/
%P 472-479
Markdown (Informal)
[Faithfulness Is Not Free: Auditing Offline KV-Cache Quantization in Retrieval-Augmented Generation](https://aclanthology.org/2026.groundlm-1.38/) (Asad et al., GroundLM 2026)
ACL