Code Generation on MBPP (Acc., TPS, Speedup)
56.8AccuracyKLASS
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| KLASSModel=Dream-v0-Instruct-7B2026.02 | 56.8 | 18.53 | 2.21 | |
| KLASSModel=Dream-v0-Base-7B2026.02 | 55.8 | 19.46 | 1.64 | |
| DAWNModel=Dream-v0-Instruct-7B2026.02 | 55.8 | 44.03 | 5.26 | |
| ConfidenceModel=Dream-v0-Instruct-7B2026.02 | 55.6 | 39.09 | 4.67 | |
| LocalLeapModel=Dream-v0-Instruct-7B2026.02 | 55 | 42.71 | 5.1 | |
| OriginalModel=Dream-v0-Instruct-7B2026.02 | 54.2 | 8.37 | 1 | |
| ConfidenceModel=Dream-v0-Base-7B2026.02 | 53.8 | 50.37 | 4.25 | |
| LocalLeapModel=Dream-v0-Base-7B2026.02 | 53.6 | 59.38 | 5.01 | |
| OriginalModel=Dream-v0-Base-7B2026.02 | 53.4 | 11.85 | 1 | |
| DAWNModel=Dream-v0-Base-7B2026.02 | 53.2 | 64.55 | 5.45 | |
| LocalLeapModel=LLaDA-1.52026.02 | 39.2 | 27 | 7.83 | |
| OriginalModel=LLaDA-1.52026.02 | 38.8 | 3.45 | 1 | |
| ConfidenceModel=LLaDA-1.52026.02 | 38.8 | 20.4 | 5.91 | |
| DAWNModel=LLaDA-1.52026.02 | 37.6 | 27.8 | 8.06 | |
| LocalLeapModel=LLaDA-8B-Instruct2026.02 | 30.8 | 44.13 | 4.66 | |
| DAWNModel=LLaDA-8B-Instruct2026.02 | 30.8 | 45.17 | 4.77 | |
| ConfidenceModel=LLaDA-8B-Instruct2026.02 | 30 | 34.81 | 3.68 | |
| KLASSModel=LLaDA-1.52026.02 | 30 | 13.25 | 3.84 | |
| OriginalModel=LLaDA-8B-Instruct2026.02 | 29.6 | 9.46 | 1 | |
| KLASSModel=LLaDA-8B-Instruct2026.02 | 27 | 22.62 | 2.39 |