List your company

Software Engineering on SWE-bench Multilingual (Pass@k)

21Pass@1SE-agent-Reflect
15.817.1518.519.85May 12, 2026

Evaluation results

MethodLinks
SE-agent-ReflectModel=Qwen-72b2026.052142.346.6
HHDModel=Qwen-72b2026.052144.348.3
Agent-RLVR-PlanModel=Qwen-72b2026.0520.441.647.3
Naive RFTModel=Qwen-72b2026.0518.342.347.7
Agent-RLVR-EFModel=Qwen-72b2026.0518.341.646.7
HHDModel=GLM-Air2026.0518.33339
Dense Expert JudgeModel=Qwen-72b2026.05184146.6
SE-agent-ReflectModel=GLM-Air2026.051830.637.3
HHD-MModel=Qwen-72b2026.0517.64145.6
HHD-MModel=GLM-Air2026.0517.33539
Agent-RLVR-EFModel=GLM-Air2026.051730.634
Dense Expert JudgeModel=GLM-Air2026.051732.338.3
Naive RFTModel=GLM-Air2026.0516.730.337.3
Agent-RLVR-PlanModel=GLM-Air2026.051631.336.3