Traces

Every run: the agent's full trajectory, its submissions over the 20-hour budget, and the final score.

Multi-GPU Efficient Finetuning

task page →5 runs · indexed 2026-09-22
#ModelRewardCostIn tokOut tokTimeStepsSubmits
1Claude Fable 50.2857$23.6614.1M128k18h 18m1192
2Claude Fable 50.2619$17.729.8M108k18h 5m1064
3Claude Fable 50.2619$30.0318.9M149k19h 3m1525
4Claude Fable 50.2381$19.0711.3M104k19h 17m1208
5Claude Fable 50.2381$17.669.8M106k18h 38m1066