About

Biography

I am currently an LLM Researcher at Alibaba Group. My research interests lie at the intersection of Reinforcement Learning (RL), Large Language Models (LLMs), and Agentic Systems.

I have been leading and contributing to several influential projects, including ROME (an open-source agentic model ecosystem) and ROLL/ROLL Flash (high-performance RL training frameworks). My goal is to build autonomous agents that can reason, self-refine, and interact effectively with complex real-world environments.

News

• [2026.02] Released a blog post on “Agentic False Positives and Environment Augmentation”.
• [2026.01] Launched ROME as part of the Agentic Learning Ecosystem (ALE).

Selected Publications

ROME: Open-source Agentic Model for Long-horizon Reasoning

Yancheng He, et al.
Technical Report / ArXiv 2026

[Agentic AI] [LLM Training]

ROLL: A Scalable Framework for Reinforcement Learning with LLMs

Yancheng He, et al.
ICLR/NeurIPS (Placeholder)

[RLHF] [Infrastructure]

Professional Services

• Reviewer: NeurIPS, ICML, ICLR
• Maintainer: iFlow CLI, ROCK Environment Engine