| Jan 09, 2026 | Today, we proudly announce the release of WebGym, the largest yet open-source RL training environment for visual web agents. The preprint can be accessed at ArXiv. We proposed (1) the RL framework with highest rollout speed, (2) recipe that supports training agents on long-horizon tasks, and (3) scaling dimensions that effectively improves the RL performance with the task set proposed. |
| Jun 11, 2025 | My first paper on web agents with RL, TTI is released! Check out the preprint! I am super proud of this work and believe it will lead to a shift of paradigm in multi-step agent reasoning with RL+VLM. |
| Jan 23, 2025 | My second paper on building device control agents with RL, Digi-Q, has been accepted to ICLR 2025! Check out the preprint! This work was done when I visited BAIR, advised by Sergey Levine and Aviral Kumar. |
| Jan 23, 2025 | My first representation learning paper CRATE-LM has been selected as Oral at CPAL 2025! This work was done when I visited BAIR, advised by Prof. Yi Ma. |
| Sep 28, 2024 | My two proud works DigiRL (first author) and RL4VLM (second author) have been accepted to NeurIPS 2024! 🎉🎉🎉 |
| Jun 20, 2024 | My first first-authored paper DigiRL is released and selected as Oral at Foundation Models in the Wild Workshop @ ICML’2024. This work is gracefully under guidance of Prof. Aviral Kumar and Prof. Sergey Levine. |
| May 15, 2024 | My first RL paper RL4VLM is released, gracefully under guidance of Prof. Sergey Levine. I’m extremely honored to also collaborate with Saining Xie and Yann Lecun. |
| Apr 12, 2024 | My first representation learning (LM pre-training) paper CRATE is accepted by JMLR! Gracefully under guidance of Prof. Yi Ma and Yaodong Yu. |
| Feb 07, 2024 | My follow-up work of SocialGen, CharmBana, has been published by WSDM’24, under guidance of Prof. Chengxiang Zhai. |
| Jun 22, 2023 | My first paper SocialGen has been published by EMNLP’23, under guidance of Prof. Chengxiang Zhai. |