Code Generation on MBPP (MB) benchmark leaderboard · SOTA2 Research

Code Generation on MBPP (MB)

65MB Score90% (CodeEnhance-gemma-4b+CodeQA) + 10% Raw

Evaluation Results

Method	Links
90% (CodeEnhance-gemma-4b+CodeQA) + 10% Raw2026.06		65
80% CodeEnhance-gemma-4b + 15% CodeQA + 5% Raw2026.06		65
95% (CodeEnhance-gemma-4b+CodeQA) + 5% Raw2026.06		64.5
85% CodeEnhance-gemma-4b + 10% CodeQA + 5% Raw2026.06		64.5
Nemotron-RW2026.06		64
75% CodeEnhance-gemma-4b + 15% CodeQA + 10% Raw2026.06		64
CodeEnhance-gemma-4b2026.06		63
95% SwallowCode + 5% Raw2026.06		58.2
50% CodeEnhance + 50% Nemotron-RW2026.06		58.2
CodeEnhance-gemma-1b2026.06		57.7
95% Nemotron-RW + 5% Raw2026.06		57.7
90% CodeEnhance+CodeQA + 10% Raw2026.06		48.2
SwallowCode2026.06		43.4
95% CodeEnhance+CodeQA + 5% Raw2026.06		36
CodeEnhance+Nemotron-RW+SwallowCode2026.06		19
CodeEnhance+CodeQA2026.06		16.4
CodeEnhance2026.06		5.6