Yongxin Deng

About

Yongxin Deng is a researcher at the forefront of integrating large language models (LLMs) with reinforcement learning (RL), focusing on how LLMs can provide structured reward guidance to improve RL task performance. Their most-cited work introduces the LMGT framework, a novel approach that leverages LLMs to generate reward functions, enabling more efficient and interpretable learning in complex environments. This contribution addresses a critical bottleneck in RL—reward design—by automating and enhancing it with semantic understanding from LLMs. With 5 citations to date, this paper has quickly gained attention for its practical implications in robotics, game AI, and autonomous systems. Deng’s research bridges natural language processing and decision-making, offering a scalable solution for tasks where traditional reward engineering is costly or infeasible. Their work is notable for its early adoption of LLMs as cognitive scaffolds in RL, a direction that promises to reshape how agents learn from human-like guidance. For students and researchers, Deng’s contributions highlight the exciting convergence of language models and reinforcement learning, opening new avenues for more adaptive and intelligent systems.

Research Focus

Key Achievements

Top Papers

1. 1
   Reward guidance for reinforcement learning tasks based on large language models: The LMGT framework

   5 citations · 2025

Key Collaborators

Contact & Links