Multi-task Language Understanding on MMLU Pro (Accuracy and Throughput)
47.12AccuracyBaseline
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BaselineModel=Dream-v0-Instruct-7B2026.03 | 47.12 | 2.01 | |
| Baseline+ParallelModel=Dream-v0-Instruct-7B2026.03 | 46.64 | 10.93 | |
| EntropyCacheModel=Dream-v0-Instruct-7B2026.03 | 45.49 | 42.14 | |
| d2CacheModel=Dream-v0-Instruct-7B2026.03 | 43.33 | 40.95 | |
| Elastic-CacheModel=Dream-v0-Instruct-7B2026.03 | 42.5 | 26.5 | |
| Fast-dLLM (Dual)Model=Dream-v0-Instruct-7B2026.03 | 41.86 | 35.01 | |
| BaselineModel=LLaDA-8B-Instruct2026.03 | 37.6 | 1.68 | |
| Baseline+ParallelModel=LLaDA-8B-Instruct2026.03 | 37.57 | 4.77 | |
| Fast-dLLM (Dual)Model=LLaDA-8B-Instruct2026.03 | 37.15 | 21.56 | |
| d2CacheModel=LLaDA-8B-Instruct2026.03 | 36.95 | 16.8 | |
| Elastic-CacheModel=LLaDA-8B-Instruct2026.03 | 36.69 | 14.78 | |
| EntropyCacheModel=LLaDA-8B-Instruct2026.03 | 36.06 | 15.21 |