Code Generation on HumanEval+ (Pass@1, AVG)

84.07Average Score (AVG)

Updated 1mo ago

Evaluation Results

Method
MT 2026.04	84.07	86.6
VCRD 2026.04	81.75	83.5
DistillLM-2 2026.04	81.37	82.3
DistilLLM 2026.04	80.75	82.3
HybKD 2026.05	47.89	36.59
Skew F(R)KL 2026.05	47.29	35.98
α-β divergence 2026.05	47.09	36.59
Forward KL 2026.05	46.76	33.54
Reverse KL 2026.05	46.34	34.15
Hard KD 2026.05	44.66	32.93
95% (CodeEnhance-gemma-4b+CodeQA) + 5% Raw 2026.06	43.3	-
Nemotron-RW 2026.06	42.1	-
90% (CodeEnhance-gemma-4b+CodeQA) + 10% Raw 2026.06	41.5	-
80% CodeEnhance-gemma-4b + 15% CodeQA + 5% Raw 2026.06	41.5	-
CodeEnhance+Nemotron-RW+SwallowCode 2026.06	41.5	-
CodeEnhance+CodeQA 2026.06	41.5	-
50% CodeEnhance + 50% Nemotron-RW 2026.06	40.9	-
85% CodeEnhance-gemma-4b + 10% CodeQA + 5% Raw 2026.06	39.6	-
SwallowCode 2026.06	39	-
75% CodeEnhance-gemma-4b + 15% CodeQA + 10% Raw 2026.06	39	-
95% CodeEnhance+CodeQA + 5% Raw 2026.06	38.4	-
95% Nemotron-RW + 5% Raw 2026.06	37.2	-
90% CodeEnhance+CodeQA + 10% Raw 2026.06	34.2	-
CodeEnhance-gemma-4b 2026.06	33.5	-
95% SwallowCode + 5% Raw 2026.06	32.3	-
CodeEnhance 2026.06	31.7	-
CodeEnhance-gemma-1b 2026.06	30.5	-