Wenhao Lin

Towards Efficient Chain-of-Thought Reasoning via Adaptive-Budgeting based Policy Optimization

Accelerating Multimodal Large Language Models via Dynamic Visual-Token Exit and the Empirical Findings