MLLMs Hallucinate when Information Distribution Drifts in Synergy Heads

Hugging Face Daily Papers Papers
multimodal-llms hallucination-reduction attention-heads causal-analysis model-trustworthiness

Summary

The paper proposes HEAL, a method to mitigate hallucinations in Multimodal Large Language Models by analyzing and calibrating information distribution in synergy heads through causal interventions and counterfactual analysis.

Multimodal Large Language Models (MLLMs) often struggle with hallucinations, thus hindering their reliable practical applications. Existing attention-based mitigation methods mainly rely on indirect signals (e.g., attention weights) that fail to accurately reflect the actual information shift underlying hallucination generation. In this paper, we propose HEAL, Head-lEvel information disentAnglement and caLibration for identifying and mitigating hallucinations. HEAL first employs causal noise intervention on multi-head outputs to filter out causally redundant heads. Subsequently, it disentangles information distribution within the remaining heads via the counterfactual Difference-in-Differences, categorizing heads into four types. Through analysis, we observe: hallucinations happen when information distribution drifts away from a healthy equilibrium in synergy heads, not strongly correlated with the quantity or strength of modality-specific heads. Motivated by this insight, HEAL injects dynamic information calibration factors into the value vectors of synergy heads, and actively regulates visual-language dependencies, steering the output distribution towards factual evidence. Extensive experiments demonstrate that HEAL effectively reduces hallucinations across multiple MLLMs, offering a simple and interpretable pathway to enhance model trustworthiness.
Original Article
View Cached Full Text

Cached at: 09/21/26, 07:20 AM

Paper page - MLLMs Hallucinate when Information Distribution Drifts in Synergy Heads

Source: https://huggingface.co/papers/2609.09206

Abstract

MultimodalLargeLanguageModels(MLLMs)oftenstrugglewithhallucinations,thushinderingtheirreliablepracticalapplications.Existingattention-basedmitigationmethodsmainlyrelyonindirectsignals(e.g.,attentionweights)thatfailtoaccuratelyreflecttheactualinformationshiftunderlyinghallucinationgeneration.Inthispaper,weproposeHEAL,Head-lEvelinformationdisentAnglementandcaLibrationforidentifyingandmitigatinghallucinations.HEALfirstemployscausalnoiseinterventiononmulti-headoutputstofilteroutcausallyredundantheads.Subsequently,itdisentanglesinformationdistributionwithintheremainingheadsviathecounterfactualDifference-in-Differences,categorizingheadsintofourtypes.Throughanalysis,weobserve:hallucinationshappenwheninformationdistributiondriftsawayfromahealthyequilibriuminsynergyheads,notstronglycorrelatedwiththequantityorstrengthofmodality-specificheads.Motivatedbythisinsight,HEALinjectsdynamicinformationcalibrationfactorsintothevaluevectorsofsynergyheads,andactivelyregulatesvisual-languagedependencies,steeringtheoutputdistributiontowardsfactualevidence.ExtensiveexperimentsdemonstratethatHEALeffectivelyreduceshallucinationsacrossmultipleMLLMs,offeringasimpleandinterpretablepathwaytoenhancemodeltrustworthiness.

View arXiv pageView PDFGitHub0Add to collection

Get this paper in your agent:

hf papers read 2609\.09206

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2609.09206 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2609.09206 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.09206 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

HalluSAE: Detecting Hallucinations in Large Language Models via Sparse Auto-Encoders

arXiv cs.CL

Researchers from Beihang University and other institutions propose HalluSAE, a framework using sparse autoencoders and phase transition theory to detect hallucinations in LLMs by modeling generation as trajectories through a potential energy landscape and identifying critical transition zones where factual errors occur.

From Architecture to Output: Structural Origins of Hallucination in Large Language Models and the Amplifying Role of Data

arXiv cs.AI

This paper analyzes hallucination in large language models as a structural consequence of three architectural decisions: self-attention's co-occurrence learning, maximum likelihood estimation training objective, and autoregressive decoding's left-to-right commitment. It maps each mechanism to specific hallucination types and argues that dataset pathologies amplify but do not cause these vulnerabilities.