Seeing without Looking: Do Vision-Language Benchmarks Really Test Vision?

• URL: http://arxiv.org/abs/2605.22903v1
• Date: Thu, 21 May 2026 17:35:04 GMT
• Title: Seeing without Looking: Do Vision-Language Benchmarks Really Test Vision?
• Abstract summary: Benchmark accuracy is often assumed to reflect grounded visual understanding in vision-language models.<n>We systematically investigate this mismatch in a set of open-source vision-language models.<n>We find that although VLMs do incorporate visual input, their predictions are less sensitive to the loss of fine-grained visual evidence.
• Score: 10.315515647818009
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Benchmark accuracy is often implicitly assumed to reflect grounded visual understanding in vision-language models (VLMs), yet it remains unclear to what extent such scores truly reflect reliance on visual evidence. Motivated by a surprising observation that removing a substantial fraction of image tokens only degrades model performance very slightly on a widely used hallucination benchmark, we systematically investigate this mismatch in a set of open-source VLMs. Our analysis spans multiple levels of granularity, spanning global visual degradation, localized occlusion, question reformulation, answer-space expansion, and decision-level analyses beyond standard accuracy. We further complement these behavioral results with a layer-wise analysis of vision-token geometry. Throughout the experiments, we find that although VLMs do incorporate visual input, their predictions are less sensitive to the loss of fine-grained visual evidence that standard accuracy should have suggested. Even when the final prediction remains unchanged, the model's internal support for the correct answer may already be weakened. We further complement a representation-level analysis, which shows increasing similarity among visual tokens in deeper layers, providing a possible explanation for our findings. Together, these results suggest that current benchmarks are not sufficient to reliably evaluate fine-grained visual grounding in VLMs.

Related papers

• Is Visual Prompting All You Need? Studying VLM Spatial Reasoning under Progressive Visual Scaffolds [12.841957617988148]
  We study how the visual presentation of a task shapes model performance and failure modes when the underlying reasoning problem is unchanged.<n>We introduce lightweight input-side scaffolds that preserve the visual modality while making spatial structure more accessible.<n>We find that visual presentation is a central factor that determines whether VLM benchmarks measure grounded perception, downstream reasoning, or a mixture of both.
  arXiv  Detail & Related papers  (2026-08-21T14:40:26Z)
• Benchmarking Large Vision-Language Models on Fine-Grained Image Tasks: From Evaluation to Diagnosis [63.49939245944781]
  We introduce FG-BMK, a comprehensive fine-grained evaluation benchmark containing 1.01 million questions and 0.28 million images.<n>We find that current LVLMs remain inadequate fine-grained recognizers, with failures arising from intertwined bottlenecks in visual representations, semantic grounding, modality alignment, and category-level knowledge.
  arXiv  Detail & Related papers  (2026-06-17T13:23:26Z)
• 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)
• Counting to Four is still a Chore for VLMs [3.017198998175514]
  Vision--language models (VLMs) have achieved impressive performance on complex multimodal reasoning tasks.<n>They still fail on simple grounding skills such as object counting.<n>We present an empirical study of VLM counting behavior through both behavioral and mechanistic analysis.
  arXiv  Detail & Related papers  (2026-04-11T05:23:19Z)
• VL-Calibration: Decoupled Confidence Calibration for Large Vision-Language Models Reasoning [5.887179947363831]
  Large Vision Language Models (LVLMs) achieve strong multimodal reasoning but frequently exhibit hallucinations and incorrect responses with high certainty.<n>Existing verbalized confidence calibration methods, largely developed for text-only LLMs, typically optimize a single holistic confidence score using binary answer-level correctness.<n>We propose VL-Calibration, a reinforcement learning framework that explicitly decouples confidence into visual and reasoning confidence.
  arXiv  Detail & Related papers  (2026-04-10T17:47:19Z)
• VLM-RobustBench: A Comprehensive Benchmark for Robustness of Vision-Language Models [64.56065206447788]
  Vision-language models (VLMs) achieve strong performance on standard, high-quality datasets.<n>We present VLM-RobustBench, a benchmark spanning 49 augmentation types across noise, blur, weather, digital, and geometrics.<n>We show that low-severity spatial perturbations often degrade performance more than visually severe photometric corruptions.
  arXiv  Detail & Related papers  (2026-03-06T10:58:02Z)
• Same Answer, Different Representations: Hidden instability in VLMs [65.36933543377346]
  We introduce a representation-aware and frequency-aware evaluation framework that measures internal embedding drift, spectral sensitivity, and structural smoothness.<n>We apply this framework to modern Vision Language Models (VLMs) across the SEEDBench, MMMU, and POPE datasets.
  arXiv  Detail & Related papers  (2026-02-06T12:24:26Z)
• ReViP: Reducing False Completion in Vision-Language-Action Models with Vision-Proprioception Rebalance [50.05984919728878]
  We present ReViP, a novel VLA framework with Vision-Proprioception Rebalance to enhance visual grounding and robustness under perturbations.<n>Specifically, we use an external VLM as a task-stage observer to extract real-time task-centric visual cues from visual observations.<n>To evaluate false completion, we propose the first False-Completion Benchmark Suite built on LIBERO with controlled settings such as Object-Drop.
  arXiv  Detail & Related papers  (2026-01-23T11:31:07Z)
• Beyond Accuracy: Evaluating Grounded Visual Evidence in Thinking with Images [34.324634481264034]
  We propose ViEBench, a process-verifiable benchmark designed to evaluate faithful visual reasoning.<n>Comprising 200 multi-scenario high-resolution images with expert-annotated visual evidence, ViEBench categorizes tasks by difficulty into perception and reasoning dimensions.<n>Our experiments yield several interesting observations: (1) VLMs can sometimes produce correct final answers despite grounding on irrelevant regions, and (2) they may successfully locate the correct evidence but still fail to utilize it to reach accurate conclusions.
  arXiv  Detail & Related papers  (2026-01-14T07:25:15Z)
• Investigate the Low-level Visual Perception in Vision-Language based Image Quality Assessment [7.969076042774561]
  We introduce a low-level distortion perception task that requires models to classify specific distortion types.<n>Our analysis shows that although MLLMs are structurally capable of representing such distortions, they tend to overfit training templates.<n>We show that improving the alignment of the vision encoder dramatically enhances distortion recognition accuracy, increasing it from 14.92% to 84.43%.
  arXiv  Detail & Related papers  (2025-12-10T12:06:47Z)
• Fooling the LVLM Judges: Visual Biases in LVLM-Based Evaluation [14.521056434373213]
  Large vision-language models (LVLMs) have emerged as the preferred tools for judging text-image alignment.<n>Can adversarial visual manipulations systematically fool LVLM judges into assigning unfairly inflated scores?<n>This work is the first study to address a key research question: Can adversarial visual manipulations systematically fool LVLM judges into assigning unfairly inflated scores?
  arXiv  Detail & Related papers  (2025-05-21T08:24:28Z)
• Detecting Multimodal Situations with Insufficient Context and Abstaining from Baseless Predictions [75.45274978665684]
  Vision-Language Understanding (VLU) benchmarks contain samples where answers rely on assumptions unsupported by the provided context.<n>We collect contextual data for each sample whenever available and train a context selection module to facilitate evidence-based model predictions.<n>We develop a general-purpose Context-AwaRe Abstention detector to identify samples lacking sufficient context and enhance model accuracy.
  arXiv  Detail & Related papers  (2024-05-18T02:21:32Z)

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.