When Text Hijacks Vision: Benchmarking and Mitigating Text Overlay-Induced Hallucination in Vision Language Models

• URL: http://arxiv.org/abs/2604.17375v1
• Date: Sun, 19 Apr 2026 10:58:40 GMT
• Title: When Text Hijacks Vision: Benchmarking and Mitigating Text Overlay-Induced Hallucination in Vision Language Models
• Abstract summary: When embedded on-screen text contradicts the visual scene, existing Vision-Language Models (VLMs) systematically hallucinate.<n>We propose VisualTextTrap, the first comprehensive benchmark, including large-scale human-validated samples.<n>We also propose Visual Text Hallucination Mitigation Mixture-of-Experts (VTHM-MoE), a novel Vision-Text Disentanglement framework.
• Score: 38.20863524973651
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Recent advances in Vision-Language Models (VLMs) have substantially enhanced their ability across multimodal video understanding benchmarks spanning temporal, action, object, and spatial understanding. However, we identify a critical yet overlooked issue: when embedded on-screen text contradicts the visual scene, existing VLMs systematically hallucinate, prioritizing overlay textual semantics over the actual visual content. We define this phenomenon as Text Overlay-Induced Hallucination (TOIH). In this work, we propose VisualTextTrap, the first comprehensive benchmark, including large-scale human-validated samples with specifically designed evaluation metrics. In particular, we construct VisualTextTrap from widely-used public datasets using a scalable hybrid pipeline of VLMs assisted text generation and rigorous manual verification. The benchmark features 6,057 samples annotated across 88 fine-grained attributes within four dimensions, with hallucination intensity quantified on a five-level scale (L1--L5) that reflects the semantic contradiction between overlay text and visual reality. Moreover, we propose Visual Text Hallucination Mitigation Mixture-of-Experts (VTHM-MoE), a novel Vision-Text Disentanglement framework that employs a dual-encoder architecture. Concretely, four dimension-specialized expert modules spanning Temporal, Action, Object, and Spatial reasoning are first pre-trained to identify and leverage cross-modal discrepancies between textual semantics and actual video content. We develop an Adaptive Token Routing Strategy to enable dynamic expert allocation, conferring robust resistance to TOIH while preserving performance on uncontaminated videos. Extensive experiments conducted on our VisualTextTrap benchmark verify the effectiveness of VTHM-MoE, outperforming state-of-the-art counterparts with diverse video question answering tasks.

Related papers

• Vision Inference Former: Sustaining Visual Consistency in Multimodal Large Language Models [58.9387276447485]
  Vision Inference Former (VIF) is a lightweight architectural module that establishes a bridge between pure visual representations and the model's output space.<n>We conduct experiments on 14 benchmark tasks covering general reasoning, OCR, table understanding, vision-centric evaluation, and hallucination.
  arXiv  Detail & Related papers  (2026-05-18T10:04:22Z)
• VISTA-Bench: Do Vision-Language Models Really Understand Visualized Text as Well as Pure Text? [51.02924254085878]
  Vision-Language Models (VLMs) have achieved impressive performance in cross-modal understanding across textual and visual inputs.<n>We introduce VISTA-Bench, a benchmark from multimodal perception, reasoning, to unimodal understanding domains.
  arXiv  Detail & Related papers  (2026-02-04T17:48:55Z)
• Revisit What You See: Disclose Language Prior in Vision Tokens for LVLM Decoding [6.612630497074871]
  Large Vision-Language Models (LVLMs) achieve strong performance across multimodal tasks by integrating visual perception with language understanding.<n>We propose ReVisiT, a training-free decoding method that references vision tokens to guide text generation.
  arXiv  Detail & Related papers  (2025-06-11T08:46:55Z)
• VidText: Towards Comprehensive Evaluation for Video Text Understanding [56.121054697977115]
  VidText is a benchmark for comprehensive and in-depth evaluation of video text understanding.<n>It covers a wide range of real-world scenarios and supports multilingual content.<n>It introduces a hierarchical evaluation framework with video-level, clip-level, and instance-level tasks.
  arXiv  Detail & Related papers  (2025-05-28T19:39:35Z)
• Text Speaks Louder than Vision: ASCII Art Reveals Textual Biases in Vision-Language Models [93.46875303598577]
  Vision-language models (VLMs) have advanced rapidly in processing multimodal information, but their ability to reconcile conflicting signals remains underexplored.<n>This work investigates how VLMs process ASCII art, a unique medium where textual elements collectively form visual patterns, potentially creating semantic-visual conflicts.
  arXiv  Detail & Related papers  (2025-04-02T10:47:07Z)
• ConTextual: Evaluating Context-Sensitive Text-Rich Visual Reasoning in Large Multimodal Models [92.60282074937305]
  We introduce ConTextual, a novel dataset featuring human-crafted instructions that require context-sensitive reasoning for text-rich images. We conduct experiments to assess the performance of 14 foundation models and establish a human performance baseline. We observe a significant performance gap of 30.8% between GPT-4V and human performance.
  arXiv  Detail & Related papers  (2024-01-24T09:07:11Z)

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.