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 2026-09-02
#ModelRewardCostIn tokOut tokTimeStepsSubmits
1DeepSeek V4 Flash Vision Exp0.3333$0.8236.7M207k19h 58m2381
2DeepSeek V4 Flash Vision Exp0.1429$0.9241.7M226k19h 50m2512
3DeepSeek V4 Flash Vision Exp0.0238$0.6629.1M168k19h 41m2262
4DeepSeek V4 Flash Vision Exp0.0000$0.7637.0M151k19h 59m2751
5DeepSeek V4 Flash Vision Exp0.0000$1.2360.8M248k19h 49m3231