Mingkang Zhu

Stratified GRPO: Handling Structural Heterogeneity in Reinforcement Learning of LLM Search Agents

From Noisy Traces to Stable Gradients: Bias--Variance Optimized Preference Optimization for Aligning Large Reasoning Models

IndustryGPT: A Large Language Model for Industrial Domain-Specific Question Answering

MiCo: Multi-image Contrast for Reinforcement Visual Reasoning

TGDPO: Harnessing Token-Level Reward Guidance for Enhancing Direct Preference Optimization

Effective LLM Knowledge Learning Requires Rethinking Generalization