通过门控感知-推理优化解决大视觉-语言模型中的过度思考
Addressing Overthinking in Large Vision-Language Models via Gated Perception-Reasoning Optimization
AI总结：
本文提出GPRO方法，通过动态路由计算路径提升大视觉-语言模型的推理效率与准确性，减少过度思考带来的冗余响应。
摘要：
大型视觉-语言模型（LVLMs）通过链式思维机制展现了强大的推理能力，但这种慢思考方法常导致过度思考，模型在简单查询中生成冗长响应，影响测试效率和准确性。本文提出门控感知-推理优化（GPRO），通过动态路由三种决策路径（轻量快速路径、慢感知路径、慢推理路径）来解决这一问题。通过大规模失败归因监督和多目标强化学习训练控制器，实验表明GPRO在五个基准测试中显著提升准确性和效率，优于近期慢思考方法并生成更短的响应。
英文摘要：
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.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
