aiqbench analysis
VisualNeedle exposes critical gaps in frontier multimodal models by demonstrating that high accuracy on standard benchmarks often relies on linguistic shortcuts rather than genuine active visual search. The study reveals that even top-tier models struggle to locate fine-grained evidence in dense scenes, with performance dropping significantly when intermediate visual tools are masked. This benchmark is essential for researchers aiming to develop MLLMs that truly utilize visual evidence rather than inferring answers from textual cues alone.
VisualNeedle: Benchmarking Active Visual Search in Information-Dense Scenes
arXiv:2605.26380v2 Announce Type: replace-cross Abstract: Frontier multimodal large language models (MLLMs) have been reported to achieve over 90\% accuracy on fine-grained perception benchmarks. However, such scores do not necessarily imply faithful use of visual evidence. Prior studies have identified three shortcuts that inflate benchmark performance. First, linguistic priors and lexical cues in questions often enable models to infer plausible answers without seeing the image. Second, coarse global semantics from the visual encoder can bypass fine-grained local details. Third, in some ``think-with-images'' benchmarks, corrupting the intermediate images returned by visual tools barely affects the final answer. These findings suggest that higher input resolution or larger question pools alone do not elicit genuine active visual search. To address this, we introduce VisualNeedle, a challenging, information-dense, and fine-grained benchmark for scenes where critical evidence is spatially constrained to minute regions and not discernible at a glance. We further propose a counterfactual crop-black setting, which replaces crops returned by tools with black images of the same size, to test whether tool-enabled performance truly relies on intermediate visual evidence.We evaluate 9 prominent MLLMs across four settings: text-only, without tools, with tools, and crop-black. Text-only accuracy stays below 10\%, while accuracy without tools remains below 20\%. The best tool-enabled model reaches only 56.00\%, still trailing the 63.00\% human majority-vote accuracy. These results reveal persistent limitations in fine-grained visual search, while the crop-black ablation confirms that success on VisualNeedle hinges on genuine intermediate visual evidence.
Abstract:Frontier multimodal large language models (MLLMs) have been reported to achieve over 90\% accuracy on fine-grained perception benchmarks. However, such scores do not necessarily imply faithful use of visual evidence. Prior studies have identified three shortcuts that inflate benchmark performance. First, linguistic priors and lexical cues in questions often enable models to infer plausible answers without seeing the image. Second, coarse global semantics from the visual encoder can bypass fine-grained local details. Third, in some ``think-with-images'' benchmarks, corrupting the intermediate images returned by visual tools barely affects the final answer. These findings suggest that higher input resolution or larger question pools alone do not elicit genuine active visual search. To address this, we introduce VisualNeedle, a challenging, information-dense, and fine-grained benchmark for scenes where critical evidence is spatially constrained to minute regions and not discernible at a glance. We further propose a counterfactual crop-black setting, which replaces crops returned by tools with black images of the same size, to test whether tool-enabled performance truly relies on intermediate visual this http URL evaluate 9 prominent MLLMs across four settings: text-only, without tools, with tools, and crop-black. Text-only accuracy stays below 10\%, while accuracy without tools remains below 20\%. The best tool-enabled model reaches only 56.00\%, still trailing the 63.00\% human majority-vote accuracy. These results reveal persistent limitations in fine-grained visual search, while the crop-black ablation confirms that success on VisualNeedle hinges on genuine intermediate visual evidence.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
arXiv:2605.26380 [cs.CV]
(or arXiv:2605.26380v2 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2605.26380 Focus to learn more arXiv-issued DOI via DataCite
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.
Full text under the publisher's open licence · 333 words
