Language Understanding on MMLU (Accuracy, Throughput, and Efficiency)
61.39AccuracyFast-dLLM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Fast-dLLMAcceleration Method=Fast-dLLM, Parallel Decoding=false2026.05 | 61.39 | 26.29 | 2.58 | |
| dKV-CacheAcceleration Method=dKV-Cache, Parallel Decoding=false2026.05 | 61.37 | 17.73 | 1.74 | |
| dLLM-CacheAcceleration Method=dLLM-Cache, Parallel Decoding=false2026.05 | 61.33 | 23.54 | 2.31 | |
| Fast-dLLMAcceleration Method=Fast-dLLM, Parallel Decoding=true2026.05 | 61.08 | 32.3 | 3.17 | |
| Dynamic-dLLMAcceleration Method=Dynamic-dLLM, Parallel Decoding=false2026.05 | 60.95 | 30.16 | 2.96 | |
| LLaDA-8B-InstructParallel Decoding=false2026.05 | 60.92 | 10.19 | 1 | |
| Dynamic-dLLMAcceleration Method=Dynamic-dLLM, Parallel Decoding=true2026.05 | 60.89 | 34.14 | 3.35 |