List your company

Code Generation on HumanEval In-Domain (Accuracy)

53.05AccuracyPion
25.146832.390939.63546.8791May 12, 2026

Evaluation results

MethodLinks
PionBase Model=Qwen2.5-1.5B2026.0553.05
AdamWBase Model=Qwen2.5-1.5B2026.0551.83
MuonBase Model=Qwen2.5-1.5B2026.0550
PionBase Model=Llama3.2-3B2026.0547.19
AdamWBase Model=Llama3.2-3B2026.0546.95
MuonBase Model=Llama3.2-3B2026.0546.34
BaseBase Model=Qwen2.5-1.5B2026.0535.98
BaseBase Model=Llama3.2-3B2026.0526.22