Agentic Critical Training (ACT) is a reinforcement learning approach that trains language‑model agents to judge actions directly by pairing expert actions with alternatives and rewarding correct selection. Unlike imitation learning or self‑reflection, ACT does not require reference rationales and can be reused across model sizes. Experiments on Qwen3‑8B and Olmo‑3‑7B‑Instruct show that ACT improves performance on ALFWorld, WebShop, ScienceWorld, and outperforms supervised reflection baselines, with significant gains over IL and IL→RL baselines.
By Weize Liu, Minghui Liu, Sy-Tuyen Ho, Yongkyun Lee, Andrew Adams Schoen, Souradip Chakraborty, Xiyao Wang, Furong Huang
arXiv:2609.38142v1 Announce Type: new
Abstract: A small trainable advisor can steer a frozen language-model executor using natural-language advice. In addition to learning from task rewards, the advi...
By Rishabh Agrawal, Hejie Cui, Shasha Li, Shanchan Wu, Sercan \"{O}. Ar{\i}k
The paper introduces Actor‑Critic with Action Chunking (AC2), a method that assigns credit to short action chunks instead of entire trajectories, enabling policy updates without waiting for terminal rewards. AC2 employs local readiness, reference solutions, and 10k‑token chunks to make critic‑based credit assignment reliable. Experiments on Qwen3‑4B with FineProofs‑RL show AC2 surpasses GRPO’s peak validation score while using 2.5× fewer decoding FLOPs and fewer training steps.
By Kaiyue Wen, Luke Bailey, Arvind Mahankali, Tengyu Ma
arXiv:2608.21501v1 Announce Type: new
Abstract: Credit assignment in large-language-model reinforcement learning (LLM RL) can be separated into three objects: evidence about success, a transport oper...
By Qifan Shi, Zhaolu Kang, Chenghua Zhu
PROOF-Gen is a method that improves distillation of tool‑calling models by recovering successful trajectories from teacher failures. It uses per‑scenario prompt optimization to generate corrective guidance that steers the teacher to a passing trajectory, then removes this guidance before training so the student learns from clean demonstrations. On τ2‑bench, PROOF-Gen recovers 93% of failed scenarios, boosting Qwen3‑4B‑Instruct‑2507’s Pass^1 from 0.132 to 0.529 and improving Gemma 4 E4B‑it by 7.2pp on BFCL v4 multi‑turn, while also raising deployed on‑device model performance by up to 5.0pp across response‑quality metrics.
By Anh Ta, Junjie Zhu, Shahin Shayandeh
arXiv:2607. 05394v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) is a powerful recipe for improving language-model reasoning, but it is expensive to repeat on every new strong model because the target model must generate many rollouts during training.
By Shiyuan Feng, Huan-ang Gao, Haohan Chi, Hanlin Wu, Zhilong Zhang, Zheng Jiang, Bingxiang He, Wei-Ying Ma, Ya-Qin Zhang, Hao Zhou
The paper introduces checkpoint handoff, an evaluation protocol that separates an agent’s ability to reach useful states from its ability to complete tasks in reinforcement learning. By using one checkpoint as a reacher up to a handoff point and another as a solver from the same replayed history, the authors can measure Reach (how often states within a fixed number of actions from success are achieved) and Solve (how often the task is completed from those states). Experiments on TravelPlanner and ALFWorld show that switching the solver from supervised fine‑tuning to RL yields larger gains when RL is used as the reacher, indicating that RL more effectively finds solvable states.
By Xuan Liu, Jingbin Qian
arXiv:2608. 07719v1 Announce Type: new Abstract: Offline reinforcement learning repeatedly trains policies from a fixed transition pool, making redundant data costly across seeds and hyperparameters, while naive subsampling can remove rare transitions needed for long-horizon credit assignment.
By Ibne Farabi Shihab, Sanjeda Akter, Abu Sa-Adat Mohamed Moon-Im Al Ahsan, Md Najmus Swaqeeb, Anuj Sharma
arXiv:2610.08402v1 Announce Type: new
Abstract: Multi-turn LLM agents often receive sparse task feedback across several interactions, while generating each response token by token. This creates two r...
By Jiaju Chen, Min Yang, Jinghua Piao, Xiaochong Lan, Xu Xia, Xiangnan He, Yong Li
Terminal‑RSI is a new framework for terminal agents that alternates Recursive Self‑Rewrite (RSR) of verified trajectories with Recursive Policy‑Refine (RPR) that inherits a critic. It introduces TB‑GAE for token‑specific baselines, Family Dense Reward (FDR) and Success‑Prioritized Reward (SPR) to calibrate progress, and uses a warm‑up critic, recursive handoff, and stage‑specific learning rates to coordinate the two refinement phases. The five‑stage pipeline applied to Qwen3.8‑27B produces Terminal‑RSI‑27B, achieving 19.84 % on Terminal‑Bench 4.0, 86.52 % on Terminal‑Bench 2.1, 16.95 % on Terminal‑Bench 3.0, and 79.00 % on Terminal‑Bench Hard, while improving process reward and strict completion on LHTB.
By Junyao Yang, Zongxia Li, Yucheng Shi, Zhongzhi Li, Ruhan Wang, Ruiwen Zhou, Haitao Mi, Shuicheng Yan, Leowei Liang
ReDraft is a reference‑driven revision method for continual post‑training of large multimodal language models. It uses the model’s own incorrect outputs as references, revises them, verifies the revisions, and fine‑tunes on the accepted ones, thereby combining explicit supervision with policy proximity. On tasks such as Counting, Clock Reading, and Jigsaw, ReDraft outperforms standard supervised fine‑tuning and on‑policy methods, achieving higher target‑task gains while dramatically reducing forgetting.
By Zhihao Zhang, Mingqi Wu, Qiaole Dong, Enyu Zhou, Shuo Li, Boyang Liu, Jiazheng Zhang, Honglin Guo, Xin Guo, Shaofan Liu, Junzhe Wang, Dingwei Zhu, Zhiheng Xi, Minlong Peng, Yuan Hua, Qi Zhang, Tao Gui, Xuanjing Huang
Reinforcement learning with verifiable rewards (RLVR) is a powerful recipe for improving language-model reasoning, but it is expensive to repeat on every new strong model because the target model must generate many rollouts during training. As models scale, post-training itself becomes a bottleneck.