Addressing Overthinking in Large Vision-Language Models via Gated Perception-Reasoning Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Diao, Xingjian, Liu, Zheyuan, Zhang, Chunhui, Wu, Weiyi, Kong, Keyi, Shi, Lin, Ding, Kaize, Vosoughi, Soroush, Gui, Jiang
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911593408757760
author Diao, Xingjian
Liu, Zheyuan
Zhang, Chunhui
Wu, Weiyi
Kong, Keyi
Shi, Lin
Ding, Kaize
Vosoughi, Soroush
Gui, Jiang
author_facet Diao, Xingjian
Liu, Zheyuan
Zhang, Chunhui
Wu, Weiyi
Kong, Keyi
Shi, Lin
Ding, Kaize
Vosoughi, Soroush
Gui, Jiang
contents Large Vision-Language Models (LVLMs) have exhibited strong reasoning capabilities through chain-of-thought mechanisms that generate step-by-step rationales. However, such slow-thinking approaches often lead to overthinking, where models produce excessively verbose responses even for simple queries, resulting in test-time inefficiency and even degraded accuracy. Prior work has attempted to mitigate this issue via adaptive reasoning strategies, but these methods largely overlook a fundamental bottleneck: visual perception failures. We argue that stable reasoning critically depends on low-level visual grounding, and that reasoning errors often originate from imperfect perception rather than insufficient deliberation. To address this limitation, we propose Gated Perception-Reasoning Optimization (GPRO), a meta-reasoning controller that dynamically routes computation among three decision paths at each generation step: a lightweight fast path, a slow perception path for re-examining visual inputs, and a slow reasoning path for internal self-reflection. To learn this distinction, we derive large-scale failure attribution supervision from approximately 790k samples, using teacher models to distinguish perceptual hallucinations from reasoning errors. We then train the controller with multi-objective reinforcement learning to optimize the trade-off between task accuracy and computational cost under uncertainty. Experiments on five benchmarks demonstrate that GPRO substantially improves both accuracy and efficiency, outperforming recent slow-thinking methods while generating significantly shorter responses.
format Preprint
id arxiv_https___arxiv_org_abs_2601_04442
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Addressing Overthinking in Large Vision-Language Models via Gated Perception-Reasoning Optimization
Diao, Xingjian
Liu, Zheyuan
Zhang, Chunhui
Wu, Weiyi
Kong, Keyi
Shi, Lin
Ding, Kaize
Vosoughi, Soroush
Gui, Jiang
Computer Vision and Pattern Recognition
Computation and Language
Large Vision-Language Models (LVLMs) have exhibited strong reasoning capabilities through chain-of-thought mechanisms that generate step-by-step rationales. However, such slow-thinking approaches often lead to overthinking, where models produce excessively verbose responses even for simple queries, resulting in test-time inefficiency and even degraded accuracy. Prior work has attempted to mitigate this issue via adaptive reasoning strategies, but these methods largely overlook a fundamental bottleneck: visual perception failures. We argue that stable reasoning critically depends on low-level visual grounding, and that reasoning errors often originate from imperfect perception rather than insufficient deliberation. To address this limitation, we propose Gated Perception-Reasoning Optimization (GPRO), a meta-reasoning controller that dynamically routes computation among three decision paths at each generation step: a lightweight fast path, a slow perception path for re-examining visual inputs, and a slow reasoning path for internal self-reflection. To learn this distinction, we derive large-scale failure attribution supervision from approximately 790k samples, using teacher models to distinguish perceptual hallucinations from reasoning errors. We then train the controller with multi-objective reinforcement learning to optimize the trade-off between task accuracy and computational cost under uncertainty. Experiments on five benchmarks demonstrate that GPRO substantially improves both accuracy and efficiency, outperforming recent slow-thinking methods while generating significantly shorter responses.
title Addressing Overthinking in Large Vision-Language Models via Gated Perception-Reasoning Optimization
topic Computer Vision and Pattern Recognition
Computation and Language
url https://arxiv.org/abs/2601.04442