IntentScore: Intent-Conditioned Action Evaluation for Computer-Use Agents 事件

IntentScore: Intent-Conditioned Action Evaluation for Computer-Use Agents arXiv:2604.05157v3 Announce Type: replace Abstract: Computer-Use Agents (CUAs) leverage large language models to execute GUI operations on desktop environments, yet they generate actions without evaluating action quality, leading to irreversible errors that cascade through subsequent steps. We propose IntentScore, a plan-aware reward model that learns to score candidate actions from 398K offline GUI interaction steps span