Mathematical Reasoning on Minerva Math 4-shot
38.4Flexibility (%)Baseline
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BaselineBackbone=Dream2026.05 | 38.4 | — | |
| P+CBackbone=Dream2026.05 | 37.7 | — | |
| DARE-KVBackbone=Dream, Base Framework=P+C2026.05 | 37.2 | 75.4 | |
| Fast-dLLMBackbone=Dream2026.05 | 37.1 | — | |
| DARE-OBackbone=Dream, Base Framework=P+C2026.05 | 36.9 | 91.5 | |
| DARE-OBackbone=Dream, Base Framework=Fast-dLLM2026.05 | 36.9 | 93.2 | |
| DARE-KVBackbone=Dream, Base Framework=Fast-dLLM2026.05 | 35.9 | 56 | |
| BaselineBackbone=LLaDA-1.52026.05 | 33.3 | — | |
| DARE-KVBackbone=LLaDA-1.5, Base Framework=P+C2026.05 | 32.5 | 38.3 | |
| P+CBackbone=LLaDA-1.52026.05 | 32 | — | |
| DARE-OBackbone=LLaDA-1.5, Base Framework=P+C2026.05 | 32 | 60.8 | |
| Fast-dLLMBackbone=LLaDA-1.52026.05 | 32 | — | |
| DARE-KVBackbone=LLaDA-1.5, Base Framework=Fast-dLLM2026.05 | 31.4 | 28.2 | |
| DARE-OBackbone=LLaDA-1.5, Base Framework=Fast-dLLM2026.05 | 30.8 | 78 | |
| DARE-KVBackbone=LLaDA, Base Framework=Fast-dLLM2026.05 | 8.6 | 28.4 | |
| Fast-dLLMBackbone=LLaDA2026.05 | 8.06 | — | |
| DARE-OBackbone=LLaDA, Base Framework=Fast-dLLM2026.05 | 8.02 | 78.1 | |
| BaselineBackbone=LLaDA2026.05 | 7.92 | — | |
| DARE-OBackbone=LLaDA, Base Framework=P+C2026.05 | 7.42 | 61.3 | |
| DARE-KVBackbone=LLaDA, Base Framework=P+C2026.05 | 7.24 | 37.9 | |
| P+CBackbone=LLaDA2026.05 | 6.94 | — |