CORTEXA
← Browse
arxivcs.CV2026-07-09

Switch-Reasoner: Learn When to Think in Multitask Mixtures via Reinforcement Learning

Yiyang Fang, Pei Fu, Jinjie Li, Jian Liang, Wenke Huang, Ruijie Luo, Shaojie Zhang, Jian Luan, Yi R. Fung, Mang Ye

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.

View free PDFSource page

Related papers

arxivcs.ROcs.CLcs.CV2026-07-31

WCM: A World Critic Model for Vision-Language-Action Reinforcement Learning

Senyu Fei, Xiaopeng Yu, Siyin Wang, Xianzhong Zhao, Jingjing Gong, Xipeng Qiu

Reinforcement learning (RL) post-training of Vision-Language-Action (VLA) models has shown strong promise for robotic manipulation. Among RL methods, critic-based approaches rely on a value estimator that predominantly operates on single-frame observations or single-frame VLM bac…

View free PDFSource page
arxivcs.AIcs.CV2026-07-23

EmoAgent-R1: Towards Multimodal Emotion Understanding with Reinforcement Learning-based Dynamic Agent Specialization

Lihuang Fang, Yuchen Zou, kebin Jin, Jinghui Qin

Multimodal large language models (MLLMs) have achieved impressive performance in multimodal emotion recognition (MER) tasks and lifted MER to a new level that is complex emotion understanding with advanced video understanding abilities and natural language description. However, e…

View free PDFSource page
arxivcs.CV2026-07-22

MoAKE: Toward Unified All-in-One Action Quality Assessment via Mixture of Action Knowledge Experts

Huangbiao Xu, Huanqi Wu, Xiao Ke, Jiaxin Cai, Junyi Wu, Jinglin Xu

Action Quality Assessment (AQA) aims to objectively evaluate performance quality from action videos. Most existing methods follow a ``one-by-one'' paradigm, training a separate model for each action type. This setting limits real-world deployment, as it requires prior action-type…

View free PDFSource page
arxivcs.ROcs.CV2026-07-23

Scale Up Strategically: Learning Compositional Generalization via Bias-Aware Evaluation and Data Collection for Robotic Manipulation

Yu Qi, Zhang Ye, Xinyi Xu, Yuxuan Lu, Amitoj Sandhu, Boce Hu, et al.

Compositional generalization is essential for robot to follow diverse instructions. However, pretrained policies are known to take shortcuts, deferring to salient cues rather than grounding language. We introduce a diagnostic framework that localizes this failure to individual \t…

View free PDFSource page
arxivcs.CV2026-07-24

Active few-shot segmentation by reinforcing data selection

Chenlan Zhao, Benny Wong, Timothy F. Lundberg, Ahmed M. Elsayed, Abdallah Aljarkas, Hamad A. Aljamaan, et al.

Few-shot learning enables medical image segmentation models to adapt to new tasks using only a small number of labelled examples. However, adaptation performance depends strongly on which examples are selected for the support set. Effective support sets should capture relevant va…

View free PDFSource page