Traces

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

SGLang Inference System Optimization

task page →5 runs · indexed 2026-09-02
#ModelRewardCostIn tokOut tokTimeStepsSubmits
1DeepSeek V4 Flash Vision Exp0.0000$2.44138.4M299k9h 52m5314
2DeepSeek V4 Flash Vision Exp0.0000$5.46332.4M492k10h 53m78215
3DeepSeek V4 Flash Vision Exp0.0000$3.59206.9M443k9h 31m59316
4DeepSeek V4 Flash Vision Exp0.0000$3.41198.5M360k8h 1m5624
5DeepSeek V4 Flash Vision Exp0.0000$3.54209.2M364k8h 32m6329