Seeing to Ground: Visual Attention for Hallucination-Resilient MDLLMs

• URL: http://arxiv.org/abs/2603.25711v1
• Date: Thu, 26 Mar 2026 17:53:49 GMT
• Title: Seeing to Ground: Visual Attention for Hallucination-Resilient MDLLMs
• Abstract summary: We introduce VISAGE, a training-free decoding framework that calibrates the objective at inference time.<n>We provide an analytical stability guarantee establishing that VISAGE maintains a bounded objective loss under estimation error.<n> Evaluations across hallucination-sensitive and general-purpose benchmarks demonstrate the robustness of the framework.
• Score: 47.94507630961399
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Multimodal Diffusion Large Language Models (MDLLMs) achieve high-concurrency generation through parallel masked decoding, yet the architectures remain prone to multimodal hallucinations. This structural vulnerability stems from an algorithmic flaw: the decoder ranks candidate tokens based on textual likelihood without verifying localized visual support. We establish that this language-only ranking induces an objective mismatch, where language probability mass acts as a misspecified proxy for the intended multimodal task. Consequently, we reinterpret hallucination as a localized optimization error, a phenomenon where the decoder exploits language shortcuts to maximize a proxy score at the expense of visual grounding. To address this objective mismatch, we introduce VISAGE, a training-free decoding framework that calibrates the objective at inference time. VISAGE estimates the proxy discrepancy by quantifying the spatial entropy of cross-attention distributions. By enforcing a localization consensus across attention heads, the method penalizes spatially uniform distributions and re-ranks token commitments to favor visually grounded outcomes. We provide an analytical stability guarantee establishing that VISAGE maintains a bounded objective loss under estimation error. Evaluations across hallucination-sensitive and general-purpose benchmarks demonstrate the robustness of the framework, yielding relative gains of 8.59% on MMMU-val and 7.75% on HallusionBench.

Related papers

• 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)
• Diagnosing Visual Ignorance in Vision-Language Models [29.2851901986069]
  Vision-Language Models (VLMs) frequently rely on language priors, producing confident answers that are weakly grounded in visual evidence.<n>In this work, we study language-prior reliance from both mechanistic and behavioral perspectives.<n>We find that language-prior reliance is a systematic routing failure affecting both model internals and benchmark validity.
  arXiv  Detail & Related papers  (2026-06-05T04:16:01Z)
• Unveiling the Visual Counting Bottleneck in Vision-Language Models [49.591496870141846]
  This work deconstructs visual counting into three cognitive stages: visual individuation, magnitude awareness, and symbolic mapping.<n>Using synthetic Go boards and linear probes, we demonstrate that visual backbones maintain robust, linearly separable representations of quantity well into the extrapolation regime.<n>We pinpoint the collapse to the symbolic mapping stage, where the model fails to project valid visual magnitudes onto symbolic tokens.
  arXiv  Detail & Related papers  (2026-05-28T16:20:29Z)
• Mitigating Object Hallucinations in Vision-Language Models through Region-Aware Attention Recalibration [7.647944993041615]
  We introduce a training-free inference strategy to dynamically correct semantic drift without relying on abrupt truncations.<n>By computing an outlier-resistant statistical midpoint across various attention heads, we establish a stable anchor for reliable visual representations.<n>This recalibration process effectively rectifies visual-semantic misalignments while fully preserving generative fluency and language priors.
  arXiv  Detail & Related papers  (2026-05-24T09:14:49Z)
• Beyond Binary Success: A Diagnostic Meta-Evaluation Framework for Fine-Grained Manipulation [98.79811866787263]
  We introduce MetaFine, a diagnostic meta-evaluation framework that disentangles manipulation competency along three axes.<n>We identify the visual encoder's ability to preserve local spatial structure as a key bottleneck for fine-grained precision.<n>By shifting evaluation from ranking to diagnosis, MetaFine turns benchmarking into an actionable compass for repairing the layered capacities underlying genuine physical dexterity.
  arXiv  Detail & Related papers  (2026-05-19T15:25:13Z)
• PDCR: Perception-Decomposed Confidence Reward for Vision-Language Reasoning [80.94559742826083]
  Reinforcement Learning with Verifiable Rewards (RLVR) traditionally relies on a sparse, outcome-based signal.<n>Recent work shows that providing a fine-grained, model-intrinsic signal effectively improves language reasoning training by providing step-level guidance without costly external models.<n>While effective for unimodal text, we find that naively applying this global reward to vision-language (V-L) reasoning is a suboptimal strategy.<n>We propose Perception-Decomposed Confidence Reward (PDCR), a framework that solves this by aligning the reward structure with the task's heterogeneous nature.
  arXiv  Detail & Related papers  (2026-05-13T12:55:18Z)
• The Cost of Context: Mitigating Textual Bias in Multimodal Retrieval-Augmented Generation [18.83395586542378]
  Multimodal Large Language Models (MLLMs) are increasingly integrated with Retrieval-Augmented Generation (RAG)<n>We identify and formalize the phenomenon of recorruption, where the introduction of even perfectly accurate "oracle" context causes a capable model to abandon an initially correct prediction.<n>We propose Bottleneck Attention Intervention for Recovery (BAIR), a parameter-free, inference-time framework that restores visual saliency and applies position-aware penalties to textual distractors.
  arXiv  Detail & Related papers  (2026-05-07T02:27:04Z)
• SAGE: Sink-Aware Grounded Decoding for Multimodal Hallucination Mitigation [33.381194425912234]
  Large vision-language models (VLMs) frequently suffer from hallucinations, generating content that is inconsistent with visual inputs.<n>We introduce SAGE, a Sink-Aware Grounded Decoding framework that mitigates hallucinations by dynamically modulating self-attention during generation.<n>Our method achieves an average relative improvement of 10.65% on MSCOCO and 7.19% on AMBER across diverse VLM architectures.
  arXiv  Detail & Related papers  (2026-03-29T22:52:03Z)
• Attention-space Contrastive Guidance for Efficient Hallucination Mitigation in LVLMs [9.043999205886658]
  Hallucinations in large vision-language models often arise when language priors dominate over visual evidence.<n>We propose Contrastive Guidance (ACG), a single-pass mechanism that operates within self-attention layers to construct both vision-language and language-only attention paths.<n>ACG achieves state-of-the-art faithfulness and caption quality while significantly reducing computational cost.
  arXiv  Detail & Related papers  (2026-01-20T08:04:18Z)
• Reasoning-Driven Amodal Completion: Collaborative Agents and Perceptual Evaluation [17.405818788700234]
  We present a Collaborative Multi-Agent Reasoning Framework that explicitly decouples Semantic Planning from Visual Synthesis.<n>Our method generates a structured, explicit plan before pixel generation, enabling visually and semantically coherent single-pass synthesis.<n>Addressing the limitations of traditional metrics in assessing inferred invisible content, we introduce the MAC-Score, a novel human-aligned evaluation metric.
  arXiv  Detail & Related papers  (2025-12-24T04:39:45Z)
• Toward More Reliable Artificial Intelligence: Reducing Hallucinations in Vision-Language Models [0.0]
  Vision-language models (VLMs) frequently generate hallucinated content plausible but incorrect claims about image content.<n>We propose a training-free self-correction framework enabling VLMs to iteratively refine responses through uncertainty-guided visual re-attention.
  arXiv  Detail & Related papers  (2025-12-08T13:58:46Z)
• Context-aware TFL: A Universal Context-aware Contrastive Learning Framework for Temporal Forgery Localization [60.73623588349311]
  We propose a universal context-aware contrastive learning framework (UniCaCLF) for temporal forgery localization.<n>Our approach leverages supervised contrastive learning to discover and identify forged instants by means of anomaly detection.<n>An efficient context-aware contrastive coding is introduced to further push the limit of instant feature distinguishability between genuine and forged instants.
  arXiv  Detail & Related papers  (2025-06-10T06:40:43Z)
• PerturboLLaVA: Reducing Multimodal Hallucinations with Perturbative Visual Training [56.172959986096316]
  This paper aims to address the challenge of hallucinations in Multimodal Large Language Models (MLLMs)<n>HalFscore is a novel metric built upon the language graph and is designed to evaluate both the accuracy and completeness of dense captions at a granular level.<n>PerturboLLaVA significantly improves the fidelity of generated captions, outperforming existing approaches in handling multimodal hallucinations.
  arXiv  Detail & Related papers  (2025-03-09T07:07:03Z)
• Transparent and Coherent Procedural Mistake Detection [30.540514590818265]
  Procedural mistake detection (PMD) is a challenging problem of classifying whether a human user has successfully executed a task (specified by a procedural text)<n>We extend PMD to require generating visual self-dialog rationales to inform decisions.<n>Given the impressive, mature image understanding capabilities observed in recent vision-and-language models (VLMs), we curate a suitable benchmark dataset for PMD based on individual frames.
  arXiv  Detail & Related papers  (2024-12-16T16:13:55Z)
• VALOR-EVAL: Holistic Coverage and Faithfulness Evaluation of Large Vision-Language Models [57.43276586087863]
  Large Vision-Language Models (LVLMs) suffer from hallucination issues, wherein the models generate plausible-sounding but factually incorrect outputs. Existing benchmarks are often limited in scope, focusing mainly on object hallucinations. We introduce a multi-dimensional benchmark covering objects, attributes, and relations, with challenging images selected based on associative biases.
  arXiv  Detail & Related papers  (2024-04-22T04:49:22Z)

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.