Reinforcement learning for computer-use agents increasingly relies on proactive reward agents that interact with the post-execution GUI environment to verify what recorded screenshots cannot reveal. The reliability of such verification hinges on two coupled decisions: what to check and when to stop. Existing methods le...
Zhi-Zhou Wang, Ya-Xi Lu, Zhong Zhang et al.· 0 citations
This work finds that finetuning a reward model to guide the policy model is more robust than directly finetuning the policy model, and proposes AgentRM, a generalizable reward model, to guide the policy model for effective test-time search.
Yu Xia, Jing-Ru Fan, Weize Chen et al.· Annual Meeting of the Associ...· 26 citations· ⚡3
Analyzing a large corpus of publicly released post-training trajectories, it is found that across different tasks, the agent's training strategy is locked in at the very beginning, and the entire remaining budget is spent on local adjustments within the selected strategy.
J. Lim, Xin Huang, Hao Peng et al.· 2 citations· ⚡1
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.