List your company

Bug Fixing on CodeAssistBench (test)

54.1AccuracyGPT-5
11.66822.68433.744.716Apr 7, 2026

Evaluation results

MethodLinks
GPT-5Interface=CodeStruct2026.0454.1122.11.1855016.74
GPT-5Interface=Baseline2026.0453.3143.91.2756619.57
GPT-5-miniInterface=CodeStruct2026.0451.983.10.896302.3
GPT-5-miniInterface=Baseline2026.0451.1125.31.266483.45
GPT-5-nanoInterface=CodeStruct2026.0448.152.81.056520.32
GPT-5-nanoInterface=Baseline2026.0446.7561.257420.34
Qwen3-CoderInterface=CodeStruct2026.0431.9779.27826363.24
Qwen3-CoderInterface=Baseline2026.0431.1829.16.95796385.61
Qwen3-32BInterface=CodeStruct2026.0420142.41.2577638.71
Qwen3-32BInterface=Baseline2026.0415.6110.52.1885829.01
Qwen3-8BInterface=CodeStruct2026.0414.10.550.179440.04
Qwen3-8BInterface=Baseline2026.0413.30.630.219840.05