Siyuan Wang

Segment-Level Attribution for Selective Learning of Long Reasoning Traces

HardcoreLogic: Challenging Large Reasoning Models with Long-tail Logic Puzzle Games

Mixture-of-Visual-Thoughts: Exploring Context-Adaptive Reasoning Mode Selection for General Visual Reasoning

Lookahead Tree-Based Rollouts for Enhanced Trajectory-Level Exploration in Reinforcement Learning with Verifiable Rewards

Not All Models Suit Expert Offloading: On Local Routing Consistency of Mixture-of-Expert Models

OViP: Online Vision-Language Preference Learning for VLM Hallucination

Autoregressive Semantic Visual Reconstruction Helps VLMs Understand Better

AutoJudger: An Agent-Driven Framework for Efficient Benchmarking of MLLMs

How Jailbreak Defenses Work and Ensemble? A Mechanistic Investigation