Wenzhe Li

Papers

Total Citations

H-Index

About

Wenzhe Li is a rising researcher at the intersection of reinforcement learning (RL) and self-supervised learning, with a focus on goal-conditioned tasks and offline decision-making. His most cited work, "Rethinking Goal-conditioned Supervised Learning and Its Connection to Offline RL" (2022, 13 citations), makes a significant theoretical contribution by unifying two previously distinct paradigms. Li demonstrates that Goal-Conditioned Supervised Learning (GCSL) can be reinterpreted through the lens of offline RL, providing a simpler, more stable alternative to traditional RL algorithms for solving sparse-reward problems. This insight not only clarifies the underlying mechanics of GCSL but also opens new avenues for applying supervised learning techniques to complex sequential decision-making tasks. By bridging self-supervised learning with offline RL, Li’s work offers practical guidance for researchers seeking to avoid the instability of value-based or policy-gradient methods. His contributions are particularly valuable for students and practitioners working on robotic manipulation, navigation, or any domain where reward signals are sparse and exploration is challenging. As an emerging voice in the field, Li’s research promises to shape how we approach goal-conditioned problems in the era of offline and self-supervised learning.