Fangzhen Lin

3 papers Department of Computer Science and Engineering, Hong Kong University of Science and Technology

• Multi-Task Vehicle Routing Solver via Mixture of Specialized Experts under State-Decomposable MDP
• Pixel Reasoner: Incentivizing Pixel Space Reasoning via Curiosity-Driven Reinforcement Learning
• VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learning