Code Generation on HumanEval (Pass Rate, CodeBLEU, Iteration Metrics) benchmark leaderboard · SOTA2 Research

Code Generation on HumanEval (Pass Rate, CodeBLEU, Iteration Metrics)

97.86Pass RateIterative

Evaluation Results

Method	Links
Iterative2026.06		97.86	1.9	0.95	0.75	0.83	0.48
Refactoring-Based2026.06		97.86	—	0.693	0.013	0.15	0.79
Iterative2026.06		97.32	1.31	0.903	0.563	0.68	0.5
No Watermark2026.06		96.28	—	0.362	0.018	0.03	0.49
Static2026.06		95.64	—	0.773	0.24	0.41	0.49
Refactoring-Based2026.06		95.63	—	0.89	0.21	0.51	0.5
ICW2026.06		94.95	—	0.74	0.01	0.12	0.49
Iterative-AGL2026.06		94	1.83	0.93	0.64	0.71	0.48
Comment-based2026.06		91.86	—	0.855	0.42	0.59	0.49
Static2026.06		90	—	0.81	0.16	0.35	0.49
Comment-based2026.06		88.9	—	0.85	0.33	0.51	0.47
Iterative-AGL2026.06		88.9	1.93	0.97	0.9	0.9	0.48
ICW2026.06		86.7	—	0.4	0.01	0.04	0.48
No Watermark2026.06		85.5	—	0.53	0.04	0.09	0.48
SWEET2026.06		26.95	—	0.81	0.45	0.54	0.31