@inproceedings{giannetti-etal-2026-mechanics,
title = "The Mechanics of Interference: Defusing Distractors in {RAG} via Sparse Autoencoder Interventions",
author = "Giannetti, Christian and
Trappolini, Giovanni and
Tonellotto, Nicola and
Silvestri, Fabrizio and
Lio, Pietro",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-acl.583/",
pages = "11999--12007",
ISBN = "979-8-89176-395-1",
abstract = "Large language models exhibit a critical vulnerability to distractor interference in retrieval-augmented contexts: they fail to prioritize relevant, factually correct documents over topically similar but misleading content. We introduce Lat-Defuse, a mechanistic framework that corrects this failure mode through targeted interventions in the model{'}s latent space. Using Sparse Autoencoders (SAEs), our method operates in an interpretable feature space and formulates correction as constrained counterfactual optimization. On Gemma-2 and Llama-3 model families across three QA benchmarks (BioASQ, Natural Questions, PopQA), our method achieves recovery rates of up to 94{\%} on distractor-vulnerable samples. Successful correction through sparse modifications reveals distractor interference as a localized, systematically addressable phenomenon, opening directions toward universal distractor robustness in LLMs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="giannetti-etal-2026-mechanics">
<titleInfo>
<title>The Mechanics of Interference: Defusing Distractors in RAG via Sparse Autoencoder Interventions</title>
</titleInfo>
<name type="personal">
<namePart type="given">Christian</namePart>
<namePart type="family">Giannetti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Giovanni</namePart>
<namePart type="family">Trappolini</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nicola</namePart>
<namePart type="family">Tonellotto</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fabrizio</namePart>
<namePart type="family">Silvestri</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pietro</namePart>
<namePart type="family">Lio</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-395-1</identifier>
</relatedItem>
<abstract>Large language models exhibit a critical vulnerability to distractor interference in retrieval-augmented contexts: they fail to prioritize relevant, factually correct documents over topically similar but misleading content. We introduce Lat-Defuse, a mechanistic framework that corrects this failure mode through targeted interventions in the model’s latent space. Using Sparse Autoencoders (SAEs), our method operates in an interpretable feature space and formulates correction as constrained counterfactual optimization. On Gemma-2 and Llama-3 model families across three QA benchmarks (BioASQ, Natural Questions, PopQA), our method achieves recovery rates of up to 94% on distractor-vulnerable samples. Successful correction through sparse modifications reveals distractor interference as a localized, systematically addressable phenomenon, opening directions toward universal distractor robustness in LLMs.</abstract>
<identifier type="citekey">giannetti-etal-2026-mechanics</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.583/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>11999</start>
<end>12007</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T The Mechanics of Interference: Defusing Distractors in RAG via Sparse Autoencoder Interventions
%A Giannetti, Christian
%A Trappolini, Giovanni
%A Tonellotto, Nicola
%A Silvestri, Fabrizio
%A Lio, Pietro
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F giannetti-etal-2026-mechanics
%X Large language models exhibit a critical vulnerability to distractor interference in retrieval-augmented contexts: they fail to prioritize relevant, factually correct documents over topically similar but misleading content. We introduce Lat-Defuse, a mechanistic framework that corrects this failure mode through targeted interventions in the model’s latent space. Using Sparse Autoencoders (SAEs), our method operates in an interpretable feature space and formulates correction as constrained counterfactual optimization. On Gemma-2 and Llama-3 model families across three QA benchmarks (BioASQ, Natural Questions, PopQA), our method achieves recovery rates of up to 94% on distractor-vulnerable samples. Successful correction through sparse modifications reveals distractor interference as a localized, systematically addressable phenomenon, opening directions toward universal distractor robustness in LLMs.
%U https://aclanthology.org/2026.findings-acl.583/
%P 11999-12007
Markdown (Informal)
[The Mechanics of Interference: Defusing Distractors in RAG via Sparse Autoencoder Interventions](https://aclanthology.org/2026.findings-acl.583/) (Giannetti et al., Findings 2026)
ACL