Look Less, Reason More: Rollout-Guided Adaptive Pixel-Space Reasoning

ACL ARR 2026 January Submission171 Authors

Keywords: Vision-Language Models, Pixel-Space Reasoning, Adaptive Thinking

Abstract: Vision-Language Models (VLMs) excel at many multimodal tasks, yet they frequently struggle with tasks requiring precise understanding and handling of fine-grained visual elements. This is mainly due to information loss during image encoding or insufficient attention to critical regions. Recent work has shown promise by incorporating pixel-level visual information into the reasoning process, enabling VLMs to access high-resolution visual details during their thought process. However, this pixel-level information is often overused, leading to inefficiency and distraction from irrelevant visual details. To address these challenges, we propose the first framework for adaptive pixel reasoning that dynamically determines necessary pixel-level operations based on the input query. Specifically, we first apply operation-aware supervised fine-tuning to establish baseline competence in textual reasoning and visual operations, then design a novel rollout-guided reinforcement learning framework relying on feedback of the model's own responses, which enables the VLM to determine when pixel operations should be invoked based on query difficulty. Experiments on extensive multimodal reasoning benchmarks show that our model achieves superior performance while significantly reducing unnecessary visual operations. Impressively, our model achieves 73.4% accuracy on HR-Bench 4K while maintaining a tool usage ratio of only 20.1%, improving accuracy and simultaneously reducing tool usage by 66.5% compared to the previous methods.

Paper Type: Long

Research Area: AI/LLM Agents

Research Area Keywords: Agentic Models, Vision-Language Models, Pixel-Space Reasoning, Adaptive Thinking

Contribution Types: Model analysis & interpretability, NLP engineering experiment

Languages Studied: English

Submission Number: 171