Multi-task Language Understanding on MMLU-Pro (Throughput and Efficiency)
45.75Throughput (TPS)LLaDA Base + dLLM-Cache
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| LLaDA Base + dLLM-CacheModel Size=8B, Caching=dLLM-Cache2025.05 | 45.75 | 3.25 | 2.15 | 3.91 | 24.69 | |
| LLaDA Instruct + dLLM-CacheModel Size=8B, Caching=dLLM-Cache2025.05 | 39.63 | 2.83 | 2.62 | 3.23 | 36.08 | |
| LLaDA BaseModel Size=8B, Caching=None2025.05 | 14.08 | 1 | 8.4 | 1 | 24.26 | |
| LLaDA InstructModel Size=8B, Caching=None2025.05 | 14.01 | 1 | 8.46 | 1 | 36.41 |