JHiMA-WebRL：分层多智能体强化学习框架提升长程网页任务表现
JHiMA-WebRL: A Joint Hierarchical Multi-Agent Reinforcement Learning Framework for Long-Horizon Web Tasks
用两个 7B 模型分工干翻 32B 单体：一个负责拆任务，一个负责执行，在 WebArena-Lite-v2 上成功率 37.2% 对 24%，做 GUI Agent 的可以看看训练细节。
论文提出 JHiMA-WebRL，一个将任务分解器与执行器耦合的分层多智能体强化学习框架，并把 GRPO 扩展到 WebArena 环境中联合优化两个模块。JHiMA-WebRL-14B（7B 执行器加 7B 分解器）在 WebArena-Lite-v2 上取得 37.2% 的平均成功率，高于同协议下用 RL 微调的 ScaleCUA-32B 的 24%。消融实验显示 RL 阶段联合更新分解器优于只更新执行器。结果表明用更小的模型做角色分工训练，可以与单体大模型在长程网页任务上竞争。
JHiMA-WebRL: A Joint Hierarchical Multi-Agent Reinforcement Learning Framework for Long-Horizon Web Tasks
Multimodal web agents have shown promising capabilities for interacting with websites through graphical user interfaces (GUIs), yet realistic web tasks often require long-horizon, multi-step planning that remains challenging for monolithic policies. Such agents must jointly infer high-level plans, navigate through intermediate states, track progress over extended interaction histories, and execute visually grounded actions. Although prior work has explored modular multi-agent systems with a separate planner, the planner is typically not updated jointly with the actor. We investigate whether explicitly separating and jointly training planning and execution provides a stronger inductive bias for long-horizon web navigation. We introduce JHiMA-WebRL, a joint hierarchical multi-agent reinforcement learning framework for web agents that couples a task decomposer with an actor. To train the decomposer, we use a strong teacher model to construct structured subtask supervision from multimodal GUI trajectories, and use these annotations to fine-tune both the decomposer and the subtask-conditioned actor. We then extend Group Relative Policy Optimization (GRPO) to jointly optimize the decomposer and actor in the interactive WebArena environment through hierarchical rollout grouping and role-specific advantages. On WebArena-Lite-v2, JHiMA-WebRL-14B, consisting of a 7B actor and a 7B decomposer, achieves an average success rate of 37.2%, compared with 24% for a ScaleCUA-32B agent that we fine-tune with RL on the same training tasks and evaluate under the same protocol. Ablations show that jointly updating the decomposer during RL outperforms updating the actor alone. These results suggest that jointly learned role factorization with smaller models can be a competitive alternative to scaling monolithic agents for long-horizon web tasks.