Saved in:
Bibliographic Details
Main Authors: Jiang, Zheng, Chen, Yiming, He, Nan, Chen, Jiahui, Li, Chaoyang, Qian, Houde, Sun, Lifeng
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.11025
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914467787309056
author Jiang, Zheng
Chen, Yiming
He, Nan
Chen, Jiahui
Li, Chaoyang
Qian, Houde
Sun, Lifeng
author_facet Jiang, Zheng
Chen, Yiming
He, Nan
Chen, Jiahui
Li, Chaoyang
Qian, Houde
Sun, Lifeng
contents Recent multimodal large language models (MLLMs) have begun to support Thinking with Images by invoking visual tools such as zooming and cropping during inference. Yet these systems remain brittle in fine-grained visual reasoning because they must decide where to look before they have access to the evidence needed to make that decision correctly. We identify this circular dependency as the Grounding Paradox. To address it, we propose Test-Time Scaling over Perception (TTSP), a framework that treats perception itself as a scalable inference process. TTSP generates multiple exploratory perception traces, filters unreliable traces using entropy-based confidence estimation, distills validated observations into structured knowledge, and iteratively refines subsequent exploration toward unresolved uncertainty. Extensive experiments on high-resolution and general multimodal reasoning benchmarks show that TTSP consistently outperforms strong baselines across backbone sizes, while also exhibiting favorable scalability and token efficiency. Our results suggest that scaling perception at test time is a promising direction for robust multimodal reasoning under perceptual uncertainty.
format Preprint
id arxiv_https___arxiv_org_abs_2604_11025
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Test-time Scaling over Perception: Resolving the Grounding Paradox in Thinking with Images
Jiang, Zheng
Chen, Yiming
He, Nan
Chen, Jiahui
Li, Chaoyang
Qian, Houde
Sun, Lifeng
Computer Vision and Pattern Recognition
Recent multimodal large language models (MLLMs) have begun to support Thinking with Images by invoking visual tools such as zooming and cropping during inference. Yet these systems remain brittle in fine-grained visual reasoning because they must decide where to look before they have access to the evidence needed to make that decision correctly. We identify this circular dependency as the Grounding Paradox. To address it, we propose Test-Time Scaling over Perception (TTSP), a framework that treats perception itself as a scalable inference process. TTSP generates multiple exploratory perception traces, filters unreliable traces using entropy-based confidence estimation, distills validated observations into structured knowledge, and iteratively refines subsequent exploration toward unresolved uncertainty. Extensive experiments on high-resolution and general multimodal reasoning benchmarks show that TTSP consistently outperforms strong baselines across backbone sizes, while also exhibiting favorable scalability and token efficiency. Our results suggest that scaling perception at test time is a promising direction for robust multimodal reasoning under perceptual uncertainty.
title Test-time Scaling over Perception: Resolving the Grounding Paradox in Thinking with Images
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2604.11025