Headroom-Drift Replay is a replay control primitive designed for GRPO that separates reuse into two decisions: Headroom ranks stored groups by remaining learning value, while Drift gates them by compatibility with the current policy. The method keeps the fresh on‑policy stream unchanged and adds no auxiliary generation or training machinery. Across mathematical reasoning, multimodal reasoning, and Agentic Search benchmarks, it outperforms naive replay and matches or exceeds broader replay methods on Avg Mean@32, delivering comparable quality at materially lower wall‑clock time in Agentic Search.
By Hyun Bin Park, Du-Seong Chang
arXiv:2606. 15333v1 Announce Type: cross Abstract: LLM unlearning has emerged as a cost-effective alternative to full retraining for removing hazardous knowledge from pretrained models while preserving general utility.
By Zirui Pang, Chenlong Zhang, Haosheng Tan, Zhuoran Jin, Jiaheng Wei, Zixin Zhong
The paper introduces a new evaluation protocol called checkpoint handoff to disentangle the contributions of reaching a target state and solving the task in reinforcement learning agents. By cloning states reached by one checkpoint and handing them to another without retraining, the authors separate the REACH metric (how often a policy arrives at a state confirmed to be a fixed number of actions from success) from the SOLVE metric (how often it finishes from that identical state). Across two benchmarks and pipelines, the analysis shows that RL history benefits RL solvers more than SFT solvers, and that independent REACH and SOLVE gaps predict overall performance.
By Xuan Liu, Jingbin Qian
arXiv:2606. 03892v1 Announce Type: cross Abstract: Training LLMs to orchestrate multi-step tool calls is held back by three coupled obstacles: realistic stateful execution environments are costly to build, synthetic training queries are often detached from the server's actual state (so the generated tool calls fail to execute), and recall-based RL rewards incentivize verbose tool-calling patterns.
By Ibrahim Abdelaziz, Asim Munawar, Kinjal Basu, Maxwell Crouse, Chulaka Gunasekara, Suneet Katrekar, Pavan Kapanipathi
arXiv:2607. 09773v1 Announce Type: new Abstract: Computer-use agents must solve long-horizon tasks through repeated interaction with partially observable, multimodal desktop environments.
By Mianqiu Huang, Taofeng Xue, Chong Peng, Jinrui Ding, Sicheng Fan, Jiale Hong, Yufei Gao, Xiaocheng Zhang, Linsen Guo, Xin Yang, Dengchang Zhao, Yuchen Xie, Peng Pei, Xunliang Xie, Xipeng Qiu
arXiv:2606. 04560v1 Announce Type: cross Abstract: Reinforcement learning from verifiable rewards with GRPO is a standard approach for post-training reasoning LLMs.
By Gyeongtae Yoo, Sanghyeok Park, Soohyuk Jang, Ik-hwan Kim, Sungroh Yoon
arXiv:2609.35954v1 Announce Type: cross
Abstract: Large language model post-training generates self-generated rollouts through reinforcement learning and on-policy distillation, yet this experience i...
By Zhiwei Zhang, Huayu Deng, Fei Zhao, Jiayan Fu, Bin Liang, Kam-Fai Wong, Mu Chuan
arXiv:2607. 21419v1 Announce Type: new Abstract: In long-horizon LLM agent reinforcement learning, weak policies often repeat similar failures, producing uninformative rollout trajectories and limiting effective policy optimization.
By Yipeng Shi, Zhipeng Ma, Yue Wang, Qitai Tan, Yang Li, Peng Chen, Zhengzhou Zhu
arXiv:2609. 04875v1 Announce Type: cross Abstract: Long-running LLM agents are stateful: beyond the transcript they accrete compressed summaries, plaintext memory, pending tool plans, and, under every serving API, a KV cache.
By Chao Yao, Yangbo Wei, Zhen Huang, Junhong Qian, Chenle Chen, Shaoqiang Lu, Chen Wu, Lei He
arXiv:2609.24174v1 Announce Type: new
Abstract: Long-horizon language agents receive sparse terminal feedback, while intermediate rubrics provide structured but potentially misspecified assessments o...
By Xuchun Hu
LEGO-RL is a framework that connects native coding-agent harnesses with scalable policy‑gradient training without altering the harnesses’ internal flow. It achieves faithful optimization through in‑process LLM proxying, reliable execution via sandbox orchestration, and observable training with automated validation and a Live UI. Experiments show LEGO‑RL improves the Qwen3.5‑35B‑A3B model’s performance on three native harnesses while preserving high rollout‑training probability correlation.
By Yiming Du, Yuxin Jiang, Tao Yuan, Jianbo Dai, Shaowei Wang, Jierun Chen, Chaofan Tao, Xianzhi Yu, Lifeng Shang, Kam-Fai Wong, Xiaohui Li, Haoli Bai
arXiv:2607. 05458v1 Announce Type: cross Abstract: Large language model (LLM) agents are usually improved by changing prompts, models, or hand-written workflows, while the execution harness around the model is treated as fixed infrastructure.
By Haiwen Yi, Xinyuan Song