Seeing without Looking: Do Vision-Language Benchmarks Really Test Vision?
This study challenges the assumption that high benchmark scores reflect true visual understanding in vision-language models (VLMs). By removing a large fraction of image tokens with minimal performance drop, the authors reveal a mismatch between accuracy and visual grounding. Through multi-level analyses including global degradation, localized occlusion, question reformulation, answer-space expansion, decision-level analysis, and layer-wise vision-token geometry, they find that models are less sensitive to fine-grained visual evidence than expected, and that visual tokens become more similar in deeper layers. The results indicate that current benchmarks are insufficient for evaluating fine-grained visual grounding in VLMs.
Article intelligence
Key points
- Removing many image tokens only slightly degrades VLM performance, questioning benchmark reliance on vision.
- Models incorporate visual input but are insensitive to loss of fine-grained visual evidence.
- Visual tokens become increasingly similar in deeper layers, offering a possible explanation.
- Current benchmarks are inadequate for assessing fine-grained visual grounding.
Why it matters
This matters because removing many image tokens only slightly degrades VLM performance, questioning benchmark reliance on vision.
Technical impact
May affect model selection, inference cost, product capability, and evaluation benchmarks.
[2605.22903] Seeing without Looking: Do Vision-Language Benchmarks Really Test Vision?
[Submitted on 21 May 2026]
Title:Seeing without Looking: Do Vision-Language Benchmarks Really Test Vision?
View a PDF of the paper titled Seeing without Looking: Do Vision-Language Benchmarks Really Test Vision?, by Zixuan Lan and 3 other authors
View PDF HTML (experimental)
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.
Comments: Accepted to GRAIL-V: Grounded Retrieval and Agentic Intelligence for Vision-Language, CVPR 2026 Workshop. accepted version
Subjects:
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
Cite as: arXiv:2605.22903 [cs.CV]
(or arXiv:2605.22903v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2605.22903
arXiv-issued DOI via DataCite (pending registration)
Submission history
From: Luzhe Sun [view email] [v1] Thu, 21 May 2026 17:35:04 UTC (19,057 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled Seeing without Looking: Do Vision-Language Benchmarks Really Test Vision?, by Zixuan Lan and 3 other authors
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.CV
new | recent | 2026-05
Change to browse by:
cs cs.AI cs.CL
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
Loading...
Data provided by:
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)