The paper introduces Continual Reasoning Gym, a continual‑RLVR environment that sequences text and visual reasoning tasks. It finds that while sequential RLVR shows modest forgetting, its final performance lags behind multitask RLVR (MTRL) because forgetting explains only part of the gap. To bridge this, the authors propose Continual Prompt Replay (CPR), which replays previous‑task prompts and regenerates responses with the current policy, achieving on average MTRL‑level performance.
By Lirui Luo, Guoxi Zhang, Hongming Xu, Rongqing Li, Cong Fang, Lifeng Fan
arXiv:2607. 04364v1 Announce Type: new Abstract: Continual post-training is becoming a central paradigm for adapting vision-language models to evolving tasks.
By Mao-Lin Luo, Zhe-Xu Wang, Zi-Hao Zhou, Bo Ye, Jian Zhao, Min-Ling Zhang, Tong Wei
arXiv:2510. 21978v2 Announce Type: replace-cross Abstract: Reinforcement learning with verifiable rewards (RLVR) has delivered impressive gains in mathematical and multimodal reasoning and has become a standard post-training paradigm for contemporary language and vision-language models.
By Hoang Phan, Xianjun Yang, Yuanshun Yao, Jingyu Zhang, Shengjie Bi, Xiaocheng Tang, Madian Khabsa, Lijuan Liu, Deren Lei
arXiv:2607. 07847v1 Announce Type: new Abstract: As large language models (LLMs) become increasingly capable, the next question is how can we enable models to continually learn?
By Anne Harrington, Nayan Saxena, Michael Murphy, Anastasia Borovykh, Zeyu Yun, Sridhar Kamath, Ara Eindra Kyi, Trevor Darrell, Jitendra Malik, Yutong Bai
arXiv:2607. 15587v1 Announce Type: new Abstract: Continual learning studies how deployed language models can continually acquire new tasks without expensive retraining from scratch.
By Yang Meng, Zhenya Liu, Zhuokai Zhao, Yuxin Chen
arXiv:2606. 17803v1 Announce Type: new Abstract: Large language models achieve strong reasoning performance by scaling inference-time compute, yet remain fundamentally stateless, discarding the rich, self-produced reasoning traces generated during this process.
By Vaggelis Dorovatas, Nancy Kalaj, Rahaf Aljundi
arXiv:2606. 11209v1 Announce Type: cross Abstract: Visual question answering increasingly requires multi-step reasoning.
By Jingpei Wu, Xiao Han, Weixiang Shen, Boer Zhang, Zifeng Ding, Volker Tresp
arXiv:2610. 22887v1 Announce Type: cross Abstract: Long rollout windows make reinforcement learning with verifiable rewards (RLVR) expensive, yet simply shortening them can remove the correctness signal: when every single-problem rollout truncates before an answer, group-relative accuracy advantages vanish.
By Bangji Yang, Hongbo Ma, Ruihan Guo, Jiajun Fan, Xi Zhu, Ge Liu
The paper introduces Harness Continual Learning (HCL), a paradigm where an agent’s state evolves through prompts, memories, tools, skills, and routing rules while keeping the underlying foundation model frozen. HCL defines harness-level forgetting and proposes a guarded evolution process involving a Continual Optimizer and Evaluator to ensure improvements without losing prior behavior. Experiments across textual reasoning, multimodal perception, and open‑world interaction show over 10% performance gains and demonstrate how the stability–plasticity trade‑off can be explicitly tuned.
By Borui Kang, Jinrui Gu, Junhan Lv, Wenbin Li, Lei Wang, Yang Gao
Large-scale online reinforcement learning (RL) is the predominant means of eliciting advanced abilities including long-term reasoning and agentic tool use in large language models (LLMs). However, continuing to scale it across vast task domains of interest remains challenging in both computational infrastructure and cost, especially when considering RL as merely a one-off learning stage.
arXiv:2607. 19450v1 Announce Type: cross Abstract: Large-scale online reinforcement learning (RL) is the predominant means of eliciting advanced abilities including long-term reasoning and agentic tool use in large language models (LLMs).
By Yunjie Chen, Xiaoxin Chen, Fang Wang
The paper demonstrates that off‑policy merging, called grafting, outperforms on‑policy self‑distillation for continual learning. Grafting learns updates on an earlier donor checkpoint, scales the update, and optionally masks sensitive directions, thereby reducing interference with existing capabilities. Across various continual learning scenarios, grafting achieves better new‑task and old‑task performance without costly on‑policy sampling.