opencua-7b-f7e0d982·2 events·first seen Aliases: OpenCUA-7B
Researchers present a systematic empirical study of inference-time scaling across four dimensions (contextual, temporal, structural, parallel) for locally-deployed computer-use agents under hardware constraints. Evaluating Qwen3-VL-8B/30B-A3B, UI-TARS-1.5-7B, and OpenCUA-7B on OSWorld, they find that additional compute often shifts rather than eliminates failure modes—contextual scaling saturates, temporal scaling extends erroneous trajectories, and structural decomposition adds overhead. The findings argue for selective compute allocation and failure-aware control mechanisms tailored to local model capabilities.
LearnWeak is an annotation-free framework for specializing small computer-use agents (CUAs) in specific software domains without deploying large expert models. It uses a stronger reference agent to identify weaknesses in a smaller student agent, synthesizes targeted tasks, and applies an error-aware training objective that disentangles planning from execution errors. On OSWorld, LearnWeak achieves gains of ~11 percentage points over 7B-8B baseline CUAs across eight domains. The work demonstrates that student-aware data synthesis substantially outperforms naive large-scale data generation for domain specialization.