Code Generation on HumanEval 256
43.9AccuracyFull
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FullModel=LLaDA 1.52026.04 | 43.9 | 256 | 44.7 | |
| ThresholdModel=LLaDA 1.52026.04 | 42.68 | 100.7 | 113.3 | |
| NI SamplingModel=LLaDA 1.52026.04 | 42.68 | 64.4 | 168.4 | |
| FullModel=LLaDA-8B-Instruct2026.04 | 37.8 | 256 | 44.1 | |
| ThresholdModel=LLaDA-8B-Instruct2026.04 | 37.2 | 98.66 | 115.8 | |
| NI SamplingModel=LLaDA-8B-Instruct2026.04 | 37.2 | 54.75 | 195.5 |