让多模态大模型学会何时思考，提升推理效率与稳定性。
Switch-Reasoner: Learn When to Think in Multitask Mixtures via Reinforcement Learning
- 用强化学习动态决定是否启用显式推理
- 在11个任务上减少无效推理，保持高准确率
- 适合需要高效推理的多任务多模态应用
多模态大语言模型通常采用固定的先思考再回答范式，在异构多任务场景中效率低下：简单输入无需推理，复杂输入则可从中获益。后训练阶段的学习也常因滚动策略不平衡导致模型趋向于始终思考或始终直接回答。我们提出Switch-Reasoner，一个基于GRPO的框架，使多模态大模型能自适应选择推理模式。该方法将思考视为虚拟工具调用，允许模型直接回答或在回答前调用显式推理。为稳定决策，引入双层调控机制，平衡整体思考与直接模式使用，并基于两种选择的相对收益提供样本级监督。在11个多模态任务上的实验表明，Switch-Reasoner减少了不必要的推理，同时维持了强性能，实现了更优的准确性-效率权衡。
原文摘要 · Abstract (English)
Multimodal Large Language Models (MLLMs) often follow a fixed Think-then-Answer paradigm, which is inefficient in heterogeneous multitask settings because simple inputs may not require explicit reasoning while difficult ones can benefit substantially from it. Learning when to think is also unstable during post-training, where imbalanced rollouts can drive the model toward always-thinking or always-direct behavior. We propose Switch-Reasoner, a GRPO-based framework that learns to adaptively select reasoning modes for MLLMs. It treats thinking as a virtual tool invocation and allows the model to either answer directly or invoke explicit reasoning before answering. To stabilize this decision, we introduce a dual-level regulation mechanism that balances the overall use of Thinking Mode and Direct Mode while providing sample-level supervision based on the relative benefit of the two choices. Experiments on 11 multimodal tasks show that Switch-Reasoner reduces unnecessary reasoning while maintaining strong performance, achieving a better accuracy-efficiency trade-off.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。