System-Mediated Attention Imbalances Make Vision-Language Models Say Yes

• URL: http://arxiv.org/abs/2601.12430v1
• Date: Sun, 18 Jan 2026 14:34:39 GMT
• Title: System-Mediated Attention Imbalances Make Vision-Language Models Say Yes
• Abstract summary: Vision-language model (VLM) hallucination is commonly linked to imbalanced allocation of attention across input modalities: system, image and text.<n>In this study, we evaluate a more holistic, system-mediated account, which attributes these imbalances to functionally redundant system weights.<n>We show that Causally redistributing attention from the system modality to image and textual inputs substantially suppresses this bias, often outperforming existing approaches.
• Score: 15.24790310403934
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Vision-language model (VLM) hallucination is commonly linked to imbalanced allocation of attention across input modalities: system, image and text. However, existing mitigation strategies tend towards an image-centric interpretation of these imbalances, often prioritising increased image attention while giving less consideration to the roles of the other modalities. In this study, we evaluate a more holistic, system-mediated account, which attributes these imbalances to functionally redundant system weights that reduce attention to image and textual inputs. We show that this framework offers a useful empirical perspective on the yes-bias, a common form of hallucination in which VLMs indiscriminately respond 'yes'. Causally redistributing attention from the system modality to image and textual inputs substantially suppresses this bias, often outperforming existing approaches. We further present evidence suggesting that system-mediated attention imbalances contribute to the yes-bias by encouraging a default reliance on coarse input representations, which are effective for some tasks but ill-suited to others. Taken together, these findings firmly establish system attention as a key factor in VLM hallucination and highlight its potential as a lever for mitigation.

Related papers

• Visual Attention Faithfulness in Vision-Language Models is Heterogeneous [13.117582974793672]
  We show that visual attention faithfulness is heterogeneous, manifesting in three distinct processing modes.<n>Human-annotated ground-truth regions satisfy comprehensiveness in only $sim 60$% of cases compared with model attention rankings.
  arXiv  Detail & Related papers  (2026-09-01T07:32:01Z)
• Disentangling Semantic Attention from Structural Bias in the Attention Manifold [61.68044165974505]
  Multimodal Large Language Models (MLLMs) exhibit disproportionate attention toward certain semantically uninformative visual tokens.<n>We introduce Saliency-guided Purification and Adaptive Redistribution (SPAR), a training-free, plug-and-play intervention.
  arXiv  Detail & Related papers  (2026-07-27T05:30:19Z)
• See Only When Needed: Context-Aware Attention Intervention for Mitigating Hallucinations in LVLMs [81.71384150441163]
  We propose Context-aware Attention Intervention (CAI) for large vision-language models (LVLMs)<n>CAI enforces a see only when needed principle via two-axis selectivity: where to look and when to intervene.<n> experiments across multiple LVLM backbones and benchmarks show that CAI achieves state-of-the-art hallucination mitigation.
  arXiv  Detail & Related papers  (2026-06-29T06:35:47Z)
• Enhancing Multimodal In-Context Learning via Inductive-Deductive Reasoning [65.15766304205657]
  In-context learning (ICL) allows large models to adapt to tasks using a few examples, yet its extension to vision-language models (VLMs) remains fragile.<n>Our analysis reveals that the fundamental limitation lies in an inductive gap, models often produce correct answers from flawed reasoning.<n>We introduce a framework that restructures multimodal ICL as a principled inductive-deductive process.
  arXiv  Detail & Related papers  (2026-05-04T09:18:19Z)
• Cognitive Pivot Points and Visual Anchoring: Unveiling and Rectifying Hallucinations in Multimodal Reasoning Models [65.57985131861399]
  hallucinations are strongly correlated with cognitive bifurcation points that often exhibit high entropy states.<n>We propose V-STAR, a lightweight, holistic training paradigm designed to internalize visually aware reasoning capabilities.
  arXiv  Detail & Related papers  (2026-04-11T13:59:05Z)
• See Fair, Speak Truth: Equitable Attention Improves Grounding and Reduces Hallucination in Vision-Language Alignment [16.616065291567445]
  Multimodal large language models (MLLMs) frequently hallucinate objects that are absent from the visual input.<n>We propose DOP-OBC, a training-free and architecture-agnostic decoding strategy built on the principle of equitable attention.
  arXiv  Detail & Related papers  (2026-04-10T11:01:48Z)
• Attention at Rest Stays at Rest: Breaking Visual Inertia for Cognitive Hallucination Mitigation [50.51650162235191]
  We propose a training-free Inertia-aware Visual Excitation (IVE) method that breaks this inertial pattern by modeling cognitive inference as the dynamic responsiveness of visual attention.<n>IVE is effective across various base MLLMs and multiple hallucination benchmarks, particularly for cognitive hallucinations.
  arXiv  Detail & Related papers  (2026-04-02T12:51:07Z)
• Mitigating Object Hallucinations in LVLMs via Attention Imbalance Rectification [23.673836358196468]
  Object hallucination in Large Vision-Language Models (LVLMs) severely compromises their reliability in real-world applications.<n>We propose Attention Imbalance Rectification (AIR), a lightweight decoding-time intervention method that reallocates attention weights and adjusts attention distributions to rectify modality-wise and token-wise imbalances.
  arXiv  Detail & Related papers  (2026-03-25T08:06:47Z)
• Revealing and Enhancing Core Visual Regions: Harnessing Internal Attention Dynamics for Hallucination Mitigation in LVLMs [67.69730908817321]
  Internal Positive Attention Dynamics (PAD) in LVLMs naturally reveal semantically core visual regions under the distortions of attention sinks.<n>We propose Positive Attention Dynamics Enhancement (PADE), a training-free attention intervention that constructs a PAD map to identify semantically core visual regions.
  arXiv  Detail & Related papers  (2026-02-17T13:08:06Z)
• Do MLLMs Really See It: Reinforcing Visual Attention in Multimodal LLMs [55.61018839017648]
  Chain-of-thought (CoT) reasoning has substantially improved multimodal large language models (MLLMs) on complex reasoning tasks.<n>Existing approaches largely rely on long textual reasoning trajectories and provide limited mechanisms for learning stable visual attention policies.<n>We propose SAYO, a visual reasoning model trained with a reinforcement learning framework that introduces a region-level visual attention-based reward.
  arXiv  Detail & Related papers  (2026-02-09T03:33:23Z)
• Seeing Through the Chain: Mitigate Hallucination in Multimodal Reasoning Models via CoT Compression and Contrastive Preference Optimization [78.94590726578014]
  multimodal reasoning models (MLRMs) remain prone to hallucinations, and effective solutions are still underexplored.<n>We propose C3PO, a training-based mitigation framework comprising textbfCompression and textbfPreference textbfOptimization.
  arXiv  Detail & Related papers  (2026-02-03T11:00:55Z)
• RoboView-Bias: Benchmarking Visual Bias in Embodied Agents for Robotic Manipulation [67.38036090822982]
  We propose RoboView-Bias, the first benchmark specifically designed to quantify visual bias in robotic manipulation.<n>We create 2,127 task instances that enable robust measurement of biases induced by individual visual factors and their interactions.<n>Our results highlight that systematic analysis of visual bias is a prerequisite for developing safe and reliable general-purpose embodied agents.
  arXiv  Detail & Related papers  (2025-09-26T13:53:25Z)
• From Bias to Balance: Exploring and Mitigating Spatial Bias in LVLMs [57.01486941224062]
  Large Vision-Language Models (LVLMs) have achieved remarkable success across a wide range of multimodal tasks.<n>We focus on how models respond when identical key visual information is placed at different locations within an image.<n>We introduce Balanced Position Assignment (BaPA), a simple yet effective mechanism that assigns identical position embeddings to all image tokens.
  arXiv  Detail & Related papers  (2025-09-26T07:07:03Z)
• Interpreting Social Bias in LVLMs via Information Flow Analysis and Multi-Round Dialogue Evaluation [1.7997395646080083]
  Large Vision Language Models (LVLMs) have achieved remarkable progress in multimodal tasks, yet they also exhibit notable social biases.<n>We propose an explanatory framework that combines information flow analysis with multi-round dialogue evaluation.<n>Experiments reveal that LVLMs exhibit systematic disparities in information usage when processing images of different demographic groups.
  arXiv  Detail & Related papers  (2025-05-27T12:28:44Z)
• More Thinking, Less Seeing? Assessing Amplified Hallucination in Multimodal Reasoning Models [43.465268635499754]
  Test-time compute has empowered large language models to generate extended reasoning chains.<n>As generations become longer, models tend to drift away from image-grounded content and rely more heavily on language priors.
  arXiv  Detail & Related papers  (2025-05-23T05:08:40Z)
• Mitigating Low-Level Visual Hallucinations Requires Self-Awareness: Database, Model and Training Strategy [53.07517728420411]
  We introduce the first instruction database specifically focused on hallucinations in low-level vision tasks.<n>We propose the Self-Awareness Failure Elimination (SAFEQA) model to improve the perception and comprehension abilities of the model in low-level vision tasks.<n>We conduct comprehensive experiments on low-level vision tasks, with the results demonstrating that our proposed method significantly enhances self-awareness of the model in these tasks and reduces hallucinations.
  arXiv  Detail & Related papers  (2025-03-26T16:05:01Z)
• Sycophancy in Vision-Language Models: A Systematic Analysis and an Inference-Time Mitigation Framework [18.54098084470481]
  We analyze sycophancy across vision-language benchmarks and propose an inference-time mitigation framework.<n>Our framework effectively mitigates sycophancy across all evaluated models, while maintaining performance on neutral prompts.
  arXiv  Detail & Related papers  (2024-08-21T01:03:21Z)
• Don't Miss the Forest for the Trees: Attentional Vision Calibration for Large Vision Language Models [16.185253476874006]
  Large Vision Language Models (LVLMs) demonstrate strong capabilities in visual understanding and description, yet often suffer from hallucinations.<n>We propose Attentional Vision (AvisC), a test-time approach that recalibrates the influence of blind tokens without modifying the underlying attention mechanism.<n>Experiments on standard benchmarks, including POPE, MME, and AMBER, demonstrate that AvisC effectively reduces hallucinations in LVLMs.
  arXiv  Detail & Related papers  (2024-05-28T04:40:57Z)
• Debiasing Multimodal Large Language Models via Penalization of Language Priors [38.97645845493758]
  Multimodal Large Language Models (MLLMs) have become indispensable tools in computer vision and natural language processing.<n>Despite their advancements, our investigation reveals a noteworthy bias: the generated content is often driven more by the inherent priors of the underlying Large Language Models (LLMs) than by the input image.<n>We propose two simple, training-free strategies to rectify these biases and redirect the model's focus toward visual information.
  arXiv  Detail & Related papers  (2024-03-08T12:35:07Z)

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.