The paper introduces Cross-Behavior Policy Optimization (CBPO), a method that unifies offline demonstrations and policy-generated rollouts into a single regression loss using a fixed-reference target. CBPO eliminates the need for separate supervised fine‑tuning and reinforcement learning objectives, allowing positive and negative offline completions, reference responses, and current and historical rollouts to share a common reward‑labeled format and replay buffer. Experiments on two model scales and four mathematical reasoning benchmarks show that CBPO, especially the CBPO‑Unified variant, outperforms baseline methods in avg@32 and pass@32 metrics, with historical rollouts providing additional gains.
By Du Cheng, Junyu Lu, Jingfan Zhang, Yuzhuo Zhang, Haoxuan Jia, Shichao Weng
arXiv:2510. 19528v2 Announce Type: replace-cross Abstract: We investigate the fundamental problem of leveraging offline data to accelerate online reinforcement learning - a direction with strong potential but limited theoretical grounding.
By Sebastian Reboul, H\'el\`ene Halconruy
arXiv:2606. 01123v1 Announce Type: new Abstract: Preference-based reinforcement learning (PbRL) avoids explicit reward engineering by learning from pairwise human preference feedback.
By Jun-Jie Yang, Chia-Heng Hsu, Kui-Yuan Chen, Ping-Chun Hsieh
The paper introduces Q-Target Pretrained Transformers (QTPT), a method that replaces supervised behavior cloning with a Bellman-style Q‑target objective for in‑context reinforcement learning. QTPT retains the context‑conditioned Transformer architecture but learns to estimate action values using rewards and transitions from the context, rather than merely imitating offline actions. The authors provide theoretical analysis in stochastic linear bandits and finite‑horizon MDPs, demonstrating improved robustness to weak or suboptimal data, and empirically show gains over supervised pretraining on controlled RL benchmarks and extensions to D4RL Kitchen and AntMaze.
By Yichen Lin, Xuyuan Xiong, Xue Wang, Xiangfu Meng, Mike Mingcheng Wei, Tao Yao
arXiv:2609.21525v1 Announce Type: new
Abstract: Preference-based reward shaping can guide reinforcement learning, but adding preference signals to the reward may unintentionally change the task being...
By Xuening Wu, Yanlan Kang, Shenqin Yin
arXiv:2609.36740v1 Announce Type: new
Abstract: Many recommender systems such as for e-commerce and news platforms aim to provide users with rankings they are likely to interact with. Off-Policy Lear...
By Ren Kishimoto, Koichi Tanaka, Haruka Kiyohara, Yusuke Narita, Yasuo Yamamoto, Nobuyuki Shimizu, Yuta Saito
arXiv:2606. 09961v1 Announce Type: cross Abstract: Training large language models (LLMs) as autonomous agents via reinforcement learning (RL) has enabled frontier models to achieve superhuman performance in long-horizon tasks.
By Yu Han, Kailing Li, Yang Jiao, Yulin Dai, Yuqian Fu, Linhai Zhuo, Tianwen Qian
arXiv:2606. 14929v1 Announce Type: cross Abstract: Modern recommendation systems increasingly rely on dynamically routing diverse queries to multiple embedding models.
By Yan Dai, Negin Golrezaei, Patrick Jaillet
arXiv:2506. 06793v2 Announce Type: replace-cross Abstract: Reward assignment from scarce demonstrations is a key challenge in both offline and online imitation learning.
By Zixuan Dong, Yumi Omori, Keith Ross
arXiv:2607. 23263v1 Announce Type: new Abstract: Deciding whether a trajectory actually fulfills its instruction governs how we measure computer-use agents on long-horizon graphical-user-interface tasks and how we train them with reinforcement learning.
By Yang Wan, Zhenhao Zhang, Jierui Wang, Linchao Zhu
arXiv:2609.13730v1 Announce Type: new
Abstract: Reliable progress in offline policy learning depends on careful reporting, well-tuned baselines, and evaluation across diverse conditions. Prior work h...
By Nabil Omi, Eric Bae, Chung Yik Edward Yeung, Siddhartha Sen, Ali Farhadi
arXiv:2602. 00781v2 Announce Type: replace Abstract: Online reinforcement learning in non-episodic, finite-horizon MDPs remains underexplored and is challenged by the need to estimate returns to a fixed terminal time.
By Jiamin Xu, Kyra Gan