List your company

Agentic Reasoning and Task Execution on Controlled-domain Multi-tool

52.8Mean Task Rubric Score (G)Base
51.44851.79952.1552.501Jul 5, 2026

Evaluation results

MethodLinks
BasePolicy Variant=Base, Seeds=3, Rollouts per task=3, Number of tasks=20, Total trials (N)=1802026.0752.8—————
Offline AW ControllerPolicy Variant=AW, Seeds=3, Rollouts per task=3, Number of tasks=20, Total trials (N)=1802026.0751.50.0130.0580.0350.2178.9