MLLMs Hallucinate when Information Distribution Drifts in Synergy Heads

• URL: http://arxiv.org/abs/2609.09206v1
• Date: Sat, 05 Sep 2026 18:01:22 GMT
• Title: MLLMs Hallucinate when Information Distribution Drifts in Synergy Heads
• Abstract summary: Multimodal Large Language Models (MLLMs) often struggle with hallucinations, thus hindering their reliable practical applications.<n>We propose HEAL, Head-lEvel information disentment and caLibration for identifying and mitigating hallucinations.
• Score: 16.959361439314616
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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.

Related papers

• Role-Break in Attention Heads: Understanding and Detecting Hallucinations in VLMs [49.743008490747094]
  Vision-Language Models (VLMs) remain prone to hallucinations, producing content that is inconsistent with or unsupported by the input image.<n>Under a unified head-level view, we find that hallucination-induced changes manifest as localized deviations from each head's faithful contextual behavior.<n> Detailed analysis reveals that these deviations are systematically organized across attention heads, contextual sources, and deviation directions, and that the resulting signal is linearly readable once head identity is preserved.
  arXiv  Detail & Related papers  (2026-07-31T13:33:51Z)
• Causal Evidence for Attention Head Imbalance in Modality Conflict Hallucination [19.688878240689423]
  Modality-conflict hallucination occurs when multimodal large language models prioritize erroneous textual premises over contradictory visual evidence.<n>We identify two groups of attention heads with opposing causal roles: hallucination-driving heads and hallucination-resisting heads.<n>Motivated by this finding, we propose MACI, a conditional intervention that suppresses causally identified hallucination-driving heads only when conflict is detected.
  arXiv  Detail & Related papers  (2026-05-19T01:47:53Z)
• VCE: A zero-cost hallucination mitigation method of LVLMs via visual contrastive editing [70.82867621856968]
  Large vision-language models (LVLMs) frequently suffer from Object Hallucination (OH)<n>Recent studies suggest that the hallucination problem may stem from language priors.<n>We propose Visual Contrastive Editing (VCE), a novel post-hoc method that identifies and suppresses hallucinatory tendencies.
  arXiv  Detail & Related papers  (2026-04-21T12:40:07Z)
• HalluGuard: Demystifying Data-Driven and Reasoning-Driven Hallucinations in LLMs [20.59139155257836]
  We introduce a unified theoretical framework that decomposes hallucination risk into data-driven and reasoning-driven components.<n>We introduce HalluGuard, an NTK-based score that leverages the induced geometry and captured representations of the NTK to jointly identify data-driven and reasoning-driven hallucinations.
  arXiv  Detail & Related papers  (2026-01-26T18:23:09Z)
• Review of Hallucination Understanding in Large Language and Vision Models [65.29139004945712]
  We present a framework for characterizing both image and text hallucinations across diverse applications.<n>Our investigations reveal that hallucinations often stem from predictable patterns in data distributions and inherited biases.<n>This survey provides a foundation for developing more robust and effective solutions to hallucinations in real-world generative AI systems.
  arXiv  Detail & Related papers  (2025-09-26T09:23:08Z)
• Exploring Causal Effect of Social Bias on Faithfulness Hallucinations in Large Language Models [50.18087419133284]
  Large language models (LLMs) have achieved remarkable success in various tasks, yet they remain vulnerable to faithfulness hallucinations.<n>We investigate whether social bias contributes to these hallucinations, a causal relationship that has not been explored.<n>A key challenge is controlling confounders within the context, which complicates the isolation of causality between bias states and hallucinations.
  arXiv  Detail & Related papers  (2025-08-11T08:34:28Z)
• MIRAGE: Assessing Hallucination in Multimodal Reasoning Chains of MLLM [58.2298313720146]
  Multimodal hallucinations are multi-sourced and arise from diverse causes.<n>Existing benchmarks fail to adequately distinguish between perception-induced hallucinations and reasoning-induced hallucinations.
  arXiv  Detail & Related papers  (2025-05-30T05:54:36Z)
• HalluciDoctor: Mitigating Hallucinatory Toxicity in Visual Instruction Data [102.56792377624927]
  hallucinations inherent in machine-generated data remain under-explored. We present a novel hallucination detection and elimination framework, HalluciDoctor, based on the cross-checking paradigm. Our method successfully mitigates 44.6% hallucinations relatively and maintains competitive performance compared to LLaVA.
  arXiv  Detail & Related papers  (2023-11-22T04:52:58Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.