List your company

Agentic Reasoning and Task Execution on Controlled-domain Coding

0.812Mean Task Rubric Score (G)Offline AW Controller
0.7080.7350.7620.789Jul 5, 2026

Evaluation results

MethodLinks
Offline AW ControllerPolicy Variant=AW, Seeds=3, Rollouts per task=3, Number of tasks=20, Total trials (N)=1802026.070.8120.10.0540.0590.0010.178
BasePolicy Variant=Base, Seeds=3, Rollouts per task=3, Number of tasks=20, Total trials (N)=1802026.070.712—————