Question Answering on GPQA (Throughput, Latency, Score)
15.04Throughputd2Cache
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| d2CacheBackbone=LLaDA-Inst, n-shot=0-shot, Generation Length=2562025.09 | 15.04 | 17.08 | 28.4 | |
| d2CacheBackbone=Dream-Inst, n-shot=0-shot, Generation Length=2562025.09 | 14.65 | 17.52 | 31.5 | |
| Fast-dLLMBackbone=LLaDA-Inst, n-shot=0-shot, Generation Length=2562025.09 | 12.41 | 20.66 | 25.7 | |
| Fast-dLLMBackbone=Dream-Inst, n-shot=0-shot, Generation Length=2562025.09 | 11.75 | 21.79 | 34.6 | |
| dLLM-CacheBackbone=LLaDA-Inst, n-shot=0-shot, Generation Length=2562025.09 | 11.51 | 22.33 | 27.2 | |
| dLLM-CacheBackbone=Dream-Inst, n-shot=0-shot, Generation Length=2562025.09 | 10.91 | 23.62 | 31 | |
| VanillaBackbone=Dream-Inst, n-shot=0-shot, Generation Length=2562025.09 | 6.43 | 41.14 | 30.1 | |
| VanillaBackbone=LLaDA-Inst, n-shot=0-shot, Generation Length=2562025.09 | 6.14 | 43.34 | 25.2 |