@inproceedings{rathore-etal-2026-robustness,
title = "Robustness of Diffusion and Autoregressive Language Models to {C}heck{L}ist-Style Perturbations",
author = "Rathore, Aarush and
Dash, Tirtharaj and
Vig, Lovekesh and
Srinivasan, Ashwin",
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-1.51/",
pages = "685--693",
ISBN = "979-8-89176-478-1",
abstract = "We compare diffusion-based and autoregressive language models under CheckList-style perturbations. We frame semantics-preserving perturbations as potentially containing an off-manifold component and hypothesise that diffusion denoising can reduce the resulting output change relative to autoregressive decoding. We evaluate eight instruction-tuned 7-9B models (three diffusion, five autoregressive) across sentiment, QQP, and SQuAD using behavioural pass rates and a perturbation coefficient that measures output change relative to input change. Results are task-dependent. Behaviourally, diffusion gains 18.2 and 11.0 points overall on QQP and SQuAD, while autoregressive models perform better on the sentiment and SQuAD invariance slices. On QQP, the all-model coefficient ratio does not favour diffusion ($\hat{γ}_{AR}/\hat{α}_{Diff} \approx 0.91); excluding two autoregressive models with strong label-prior effects as a diagnostic sensitivity analysis yields 1.23 overall and 1.48 on invariance tests. Diffusion also performs worse on perturbations that should change the output. Thus, the observed stability differences are task-conditional, and because the empirical diffusion coefficients exceed one, the experiments do not directly validate the contraction assumption of our theoretical proposition. Data and code are available at: https://github.com/AarushRathore1729/dlm_ar_final.$"
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="rathore-etal-2026-robustness">
<titleInfo>
<title>Robustness of Diffusion and Autoregressive Language Models to CheckList-Style Perturbations</title>
</titleInfo>
<name type="personal">
<namePart type="given">Aarush</namePart>
<namePart type="family">Rathore</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tirtharaj</namePart>
<namePart type="family">Dash</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lovekesh</namePart>
<namePart type="family">Vig</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ashwin</namePart>
<namePart type="family">Srinivasan</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>We compare diffusion-based and autoregressive language models under CheckList-style perturbations. We frame semantics-preserving perturbations as potentially containing an off-manifold component and hypothesise that diffusion denoising can reduce the resulting output change relative to autoregressive decoding. We evaluate eight instruction-tuned 7-9B models (three diffusion, five autoregressive) across sentiment, QQP, and SQuAD using behavioural pass rates and a perturbation coefficient that measures output change relative to input change. Results are task-dependent. Behaviourally, diffusion gains 18.2 and 11.0 points overall on QQP and SQuAD, while autoregressive models perform better on the sentiment and SQuAD invariance slices. On QQP, the all-model coefficient ratio does not favour diffusion (\hatγ_AR/\hatα_Diff \approx 0.91); excluding two autoregressive models with strong label-prior effects as a diagnostic sensitivity analysis yields 1.23 overall and 1.48 on invariance tests. Diffusion also performs worse on perturbations that should change the output. Thus, the observed stability differences are task-conditional, and because the empirical diffusion coefficients exceed one, the experiments do not directly validate the contraction assumption of our theoretical proposition. Data and code are available at: https://github.com/AarushRathore1729/dlm_ar_final.</abstract>
<identifier type="citekey">rathore-etal-2026-robustness</identifier>
<location>
<url>https://aclanthology.org/2026.groundlm-1.51/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>685</start>
<end>693</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Robustness of Diffusion and Autoregressive Language Models to CheckList-Style Perturbations
%A Rathore, Aarush
%A Dash, Tirtharaj
%A Vig, Lovekesh
%A Srinivasan, Ashwin
%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 rathore-etal-2026-robustness
%X We compare diffusion-based and autoregressive language models under CheckList-style perturbations. We frame semantics-preserving perturbations as potentially containing an off-manifold component and hypothesise that diffusion denoising can reduce the resulting output change relative to autoregressive decoding. We evaluate eight instruction-tuned 7-9B models (three diffusion, five autoregressive) across sentiment, QQP, and SQuAD using behavioural pass rates and a perturbation coefficient that measures output change relative to input change. Results are task-dependent. Behaviourally, diffusion gains 18.2 and 11.0 points overall on QQP and SQuAD, while autoregressive models perform better on the sentiment and SQuAD invariance slices. On QQP, the all-model coefficient ratio does not favour diffusion (\hatγ_AR/\hatα_Diff \approx 0.91); excluding two autoregressive models with strong label-prior effects as a diagnostic sensitivity analysis yields 1.23 overall and 1.48 on invariance tests. Diffusion also performs worse on perturbations that should change the output. Thus, the observed stability differences are task-conditional, and because the empirical diffusion coefficients exceed one, the experiments do not directly validate the contraction assumption of our theoretical proposition. Data and code are available at: https://github.com/AarushRathore1729/dlm_ar_final.
%U https://aclanthology.org/2026.groundlm-1.51/
%P 685-693
Markdown (Informal)
[Robustness of Diffusion and Autoregressive Language Models to CheckList-Style Perturbations](https://aclanthology.org/2026.groundlm-1.51/) (Rathore et al., GroundLM 2026)
ACL