Efficient Agent Training for Computer Use
This work addresses the data efficiency problem for researchers and developers training AI agents for computer use, representing a strong specific gain rather than a broad paradigm shift.
The paper tackles the bottleneck of scaling high-quality trajectory data for computer use agents by introducing PC Agent-E, which achieved a 141% relative improvement over Claude 3.7 Sonnet on the WindowsAgentArena-V2 benchmark using only 312 human-annotated trajectories enhanced with synthetic data.
Scaling up high-quality trajectory data has long been a critical bottleneck for developing human-like computer use agents. We introduce PC Agent-E, an efficient agent training framework that significantly reduces reliance on large-scale human demonstrations. Starting with just 312 human-annotated computer use trajectories, we further improved data quality by synthesizing diverse action decisions with Claude 3.7 Sonnet. Trained on these enriched trajectories, our PC Agent-E model achieved a remarkable 141% relative improvement, surpassing the strong Claude 3.7 Sonnet with extended thinking on WindowsAgentArena-V2, an improved benchmark we also released. Furthermore, PC Agent-E demonstrates strong generalizability to different operating systems on OSWorld. Our findings suggest that strong computer use capabilities can be stimulated from a small amount of high-quality trajectory data.