Search papers, labs, and topics across Lattice.
This paper introduces LearnWeak, a framework for specializing small computer-use agents (CUAs) by leveraging a stronger reference agent to identify weaknesses in specific domains. LearnWeak synthesizes targeted training tasks and constructs supervision automatically, using an error-aware objective to differentiate planning and execution errors. Experiments on OSWorld show LearnWeak improves EvoCUA-8B and OpenCUA-7B performance by over 11 percentage points across eight domains, demonstrating the value of student-aware data synthesis and training.
Naively scaling training data for small computer-use agents yields marginal improvements, but targeting training on weaknesses identified by a stronger agent yields substantial gains.
Computer-use agents (CUAs) have recently made substantial progress, but deploying a separate large expert for each software domain remains expensive. Small open computer-use agents are more practical specialization targets, but they remain substantially weaker and exhibit uneven domain-specific failures. A straightforward remedy is to synthesize large-scale training data for the target domain, yet we find that this naive approach yields only marginal improvements. Building on this observation, we introduce LearnWeak, an annotation-free specialization framework for small computer-use agents that uses a stronger reference agent to identify the student's weaknesses in the target domain, synthesize targeted tasks, and construct supervision automatically. LearnWeak further introduces an error-aware specialization objective that disentangles planning and execution errors, enabling more behaviorally precise updates than broad uniform supervision. On OSWorld, LearnWeak achieves average gains of 11.6 and 11.1 percentage points over EvoCUA-8B and OpenCUA-7B, respectively, across eight domains. We also validate that our student-aware dataset generation and training approaches outperform existing autonomous trajectory generation and training baselines. Our work highlights the importance of student awareness in both data synthesis and agent training, pointing toward a more principled and efficient path for specializing small computer-use agents in diverse domains.
