Jiaqi WANG

On the Measurement and Efficient Mitigation of Length Generalization Gaps in Large Language Models

A Signed Graph Approach to Understanding and Mitigating Oversmoothing

Visual Thoughts: A Unified Perspective of Understanding Multimodal Chain-of-Thought

Think or Not? Selective Reasoning via Reinforcement Learning for Vision-Language Models

PIGDreamer: Privileged Information Guided World Models for Safe Partially Observable Reinforcement Learning

Oversmoothing as Loss of Sign: Towards Structural Balance in Graph Neural Networks