List your company

Code Generation on HumanEval+, MBPP+, LCB

85.2HumanEval+ ScoreTeacher
60.96867.25973.5579.841May 5, 2026

Evaluation results

MethodLinks
TeacherModel Type=Teacher (RL), Teacher Models=Qwen3-4B-Math-RL and Qwen3-4B-Code-RL2026.0585.269.826.660.5
Uni-OPDStudent Model=Qwen3-4B2026.0583.171.32860.8
OPDStudent Model=Qwen3-4B2026.0582.972.421.659
TeacherModel=Qwen3-30B-A3B-Instruct-25072026.0581.977.223.460.8
StudentStudent Model=Qwen3-4B2026.0577.465.317.753.5
OPDDistillation Setting=Multi-Teacher Distillation2026.0573.259.125.752.7
Uni-OPDDistillation Setting=Single-Teacher Distillation2026.05736028.153.7
Uni-OPDDistillation Setting=Multi-Teacher Distillation2026.0572.960.52853.8
OPDDistillation Setting=Single-Teacher Distillation2026.0571.858.226.752.5
Uni-OPDStudent Model=Qwen3-1.7B2026.0571.558.62852.7
OPDStudent Model=Qwen3-1.7B2026.0567.156.723.449.1
StudentModel Type=Qwen3-1.7B Student2026.0561.953.411.942.4
StudentStudent Model=Qwen3-1.7B2026.0561.953.411.942.4