Yanzhe Hu
CV

LightningRL framework showing policy optimization, sampling and reward, and block-wise diffusion language model inference before and after training.
ICML,2026First author

LightningRL: Breaking the Accuracy-Parallelism Trade-off of Block-wise dLLMs via Reinforcement Learning

Yanzhe Hu, Yijie Jin, Pengfei Liu, Kai Yu, Zhijie Deng

A reinforcement learning framework for improving the inference efficiency of block-wise diffusion language models.

WebPaperCode
Diffusion language models · Reinforcement learning
Representative dependency-tree topologies and interconnection statistics from the large language model software supply-chain study.
Internetware,2026First author

Understanding Large Language Model Supply Chain: Structure, Domain, and Vulnerabilities

Yanzhe Hu*, Shenao Wang*, Tianyuan Nie, Yanjie Zhao, Haoyu Wang

A large-scale empirical study of the structure and security risks of the LLM software supply chain.

Paper
LLM supply chains · Software security
World-Language-Action architecture with action and world experts, physical dynamics modeling, and value-based trajectory selection.
CoRL,2026Submitted

World-Language-Action Model for Unified World Modeling, Language Reasoning, and Action Synthesis

Yi Yang, Zhihong Liu, Siqi Kou, Yiyang Chen, Yanzhe Hu, et al.

A unified model connecting world modeling, language reasoning, and action synthesis.

PaperCode
World models · Embodied AI