@inproceedings{movva-loke-2026-quantization,
title = "Does Quantization Erode Grounding? A Faithfulness Audit of On-Device {LLM} Compression",
author = "Movva, Neeraj Sai Venkat and
Loke, Sathvik",
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-1.16/",
pages = "195--200",
ISBN = "979-8-89176-478-1",
abstract = "Weight quantization is the default route to running language models on consumer hardware, yet its effect on contextual faithfulness{---}a model{'}s willingness to ground answers in provided context rather than memorized priors{---}remains untested: faithfulness benchmarks use full-precision models, and quantization studies measure accuracy and perplexity, not grounding. We audit two instruction-tuned 3B models (Qwen2.5-3B, Llama-3.2-3B) across seven GGUF precisions from 16-bit to 2-bit, under aligned, counterfactual, and unanswerable contexts, using 300 SQuAD-derived items and three prompt templates per condition (900 trials per condition per precision). With an item-level clustered bootstrap we find faithfulness unchanged from 16 down to 5 bits in both models, while below that small but statistically detectable declines appear whose onset is model-dependent (4-bit for Llama, 3-bit for Qwen). Critically, production of the original, context-contradicting answer does not rise at any precision, so the intuitive ``quantization erodes grounding'' mechanism is not supported: lost faithfulness moves to off-target responses instead. Qwen additionally collapses into incoherent output at 2-bit while Llama stays fluent, making extreme compression a per-model question. Code and harness: https://github.com/movvaneeraj/quantfaith."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="movva-loke-2026-quantization">
<titleInfo>
<title>Does Quantization Erode Grounding? A Faithfulness Audit of On-Device LLM Compression</title>
</titleInfo>
<name type="personal">
<namePart type="given">Neeraj</namePart>
<namePart type="given">Sai</namePart>
<namePart type="given">Venkat</namePart>
<namePart type="family">Movva</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sathvik</namePart>
<namePart type="family">Loke</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>Weight quantization is the default route to running language models on consumer hardware, yet its effect on contextual faithfulness—a model’s willingness to ground answers in provided context rather than memorized priors—remains untested: faithfulness benchmarks use full-precision models, and quantization studies measure accuracy and perplexity, not grounding. We audit two instruction-tuned 3B models (Qwen2.5-3B, Llama-3.2-3B) across seven GGUF precisions from 16-bit to 2-bit, under aligned, counterfactual, and unanswerable contexts, using 300 SQuAD-derived items and three prompt templates per condition (900 trials per condition per precision). With an item-level clustered bootstrap we find faithfulness unchanged from 16 down to 5 bits in both models, while below that small but statistically detectable declines appear whose onset is model-dependent (4-bit for Llama, 3-bit for Qwen). Critically, production of the original, context-contradicting answer does not rise at any precision, so the intuitive “quantization erodes grounding” mechanism is not supported: lost faithfulness moves to off-target responses instead. Qwen additionally collapses into incoherent output at 2-bit while Llama stays fluent, making extreme compression a per-model question. Code and harness: https://github.com/movvaneeraj/quantfaith.</abstract>
<identifier type="citekey">movva-loke-2026-quantization</identifier>
<location>
<url>https://aclanthology.org/2026.groundlm-1.16/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>195</start>
<end>200</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Does Quantization Erode Grounding? A Faithfulness Audit of On-Device LLM Compression
%A Movva, Neeraj Sai Venkat
%A Loke, Sathvik
%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 movva-loke-2026-quantization
%X Weight quantization is the default route to running language models on consumer hardware, yet its effect on contextual faithfulness—a model’s willingness to ground answers in provided context rather than memorized priors—remains untested: faithfulness benchmarks use full-precision models, and quantization studies measure accuracy and perplexity, not grounding. We audit two instruction-tuned 3B models (Qwen2.5-3B, Llama-3.2-3B) across seven GGUF precisions from 16-bit to 2-bit, under aligned, counterfactual, and unanswerable contexts, using 300 SQuAD-derived items and three prompt templates per condition (900 trials per condition per precision). With an item-level clustered bootstrap we find faithfulness unchanged from 16 down to 5 bits in both models, while below that small but statistically detectable declines appear whose onset is model-dependent (4-bit for Llama, 3-bit for Qwen). Critically, production of the original, context-contradicting answer does not rise at any precision, so the intuitive “quantization erodes grounding” mechanism is not supported: lost faithfulness moves to off-target responses instead. Qwen additionally collapses into incoherent output at 2-bit while Llama stays fluent, making extreme compression a per-model question. Code and harness: https://github.com/movvaneeraj/quantfaith.
%U https://aclanthology.org/2026.groundlm-1.16/
%P 195-200
Markdown (Informal)
[Does Quantization Erode Grounding? A Faithfulness Audit of On-Device LLM Compression](https://aclanthology.org/2026.groundlm-1.16/) (Movva & Loke, GroundLM 2026)
ACL