Regularized Score Centeringwith Chenlu Ye, Xuanchang Zhang, Aviral Kumar, Tong Zhang1 versionW_x5k2b8aqv1 · current
Enabling Token-wise Importance Sampling in Off-Policy RLVR with Proximal KL RegularizationHao Bai, Jiarui Yao, Tong Zhang0 citations2610.18738v12026arXiv
FARS: A Fully Automated Research System Deployed at ScaleQiong Tang, Tianxiang Sun, Xiangkun Hu et al.0 citations2606.31651v22026arXiv
OpenWebRL: Demystifying Online Multi-turn Reinforcement Learning for Visual Web AgentsRui Yang, Qianhui Wu, Yuxi Chen et al.0 citations2606.02031v22026arXiv
InT: Self-Proposed Interventions Enable Credit Assignment in LLM ReasoningMatthew Y. R. Yang, Hao Bai, Ian Wu et al.0 citations2601.14209v12026arXiv
Future-KL Regularized GRPO: Process-Level Credit Assignment from fff-Divergence RegularizationJiarui Yao, Ruida Wang, Hao Bai et al.0 citations2601.10201v22026arXiv