Code Generation on HumanEval (Pass@1, ET, Step, Time)
45.12Pass@1Saber
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SaberSampling Strategy=Efficient DLM Sampling, Backbone=LLaDA-8B-Instruct2025.10 | 45.12 | 0.3598 | 118.92 | 41 | |
| ConfidenceSampling Strategy=Standard DLM Sampling, Backbone=LLaDA-8B-Instruct2025.10 | 43.29 | 0.3579 | 256 | 2 | |
| EntropySampling Strategy=Standard DLM Sampling, Backbone=LLaDA-8B-Instruct2025.10 | 41.46 | 0.3415 | 256 | 1 | |
| WINOSampling Strategy=Efficient DLM Sampling, Backbone=LLaDA-8B-Instruct2025.10 | 40.24 | 0.3171 | 100.12 | 57 | |
| Fast-dLLMSampling Strategy=Efficient DLM Sampling, Backbone=LLaDA-8B-Instruct2025.10 | 39.63 | 0.3415 | 256 | 59 | |
| Fast-dLLM (+parallel)Sampling Strategy=Efficient DLM Sampling, Parallelism=true, Backbone=LLaDA-8B-Instruct2025.10 | 39.63 | 0.3354 | 96.24 | 25 | |
| SAR (p=2)Sampling Strategy=Efficient DLM Sampling, p parameter=2, Backbone=LLaDA-8B-Instruct2025.10 | 35.98 | 0.2927 | 128 | 1 | |
| Confidence (p=2)Sampling Strategy=Efficient DLM Sampling, p parameter=2, Backbone=LLaDA-8B-Instruct2025.10 | 34.76 | 0.2866 | 128 | 51 | |
| ReMDMSampling Strategy=Efficient DLM Sampling, Backbone=LLaDA-8B-Instruct2025.10 | 20.73 | 0.1829 | 128 | 1 | |
| RandomSampling Strategy=Standard DLM Sampling, Backbone=LLaDA-8B-Instruct2025.10 | 14.63 | 0.128 | 256 | 1 |