Mathematical Reasoning on GSM8K (test) (Accuracy, TPS)
83.02AccuracyFDM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FDMBackbone=LLaDA-1.5, Search Width (K)=42025.12 | 83.02 | 4.69 | |
| FDMBackbone=LLaDA-1.5, Search Width (K)=32025.12 | 82.64 | 5.84 | |
| FDMBackbone=LLaDA-1.5, Search Width (K)=22025.12 | 82.49 | 8.29 | |
| FDMBackbone=LLaDA-8B-Instruct, Search Width (K)=42025.12 | 82.34 | 4.61 | |
| FDMBackbone=LLaDA-8B-Instruct, Search Width (K)=32025.12 | 82.18 | 5.86 | |
| MarginBackbone=LLaDA-1.5, Step Number (T)=2562025.12 | 82.18 | 11.39 | |
| ProbabilityBackbone=LLaDA-1.5, Step Number (T)=2562025.12 | 82.1 | 11.89 | |
| FDMBackbone=LLaDA-8B-Instruct, Search Width (K)=22025.12 | 82.03 | 8.28 | |
| EntropyBackbone=LLaDA-1.5, Step Number (T)=2562025.12 | 81.93 | 10.94 | |
| ProbabilityBackbone=LLaDA-8B-Instruct, Step Number (T)=2562025.12 | 81.2 | 11.51 | |
| MarginBackbone=LLaDA-8B-Instruct, Step Number (T)=2562025.12 | 80.14 | 11.23 | |
| EntropyBackbone=LLaDA-8B-Instruct, Step Number (T)=2562025.12 | 80.12 | 10.79 | |
| FDMBackbone=LLaDA-MoE-7B-Instruct, Search Width (K)=42025.12 | 78.32 | 3.6 | |
| FDMBackbone=LLaDA-MoE-7B-Instruct, Search Width (K)=32025.12 | 77.63 | 3.78 | |
| FDMBackbone=LLaDA-MoE-7B-Instruct, Search Width (K)=22025.12 | 77.48 | 3.8 | |
| MarginBackbone=LLaDA-MoE-7B-Instruct, Step Number (T)=2562025.12 | 76.8 | 4.04 | |
| EntropyBackbone=LLaDA-MoE-7B-Instruct, Step Number (T)=2562025.12 | 76.8 | 3.84 | |
| ProbabilityBackbone=LLaDA-MoE-7B-Instruct, Step Number (T)=2562025.12 | 76.5 | 4.13 | |
| FDMBackbone=MMaDA-8B-MixCoT, Search Width (K)=42025.12 | 57.92 | 4.61 | |
| FDMBackbone=MMaDA-8B-MixCoT, Search Width (K)=32025.12 | 57.68 | 5.71 | |
| FDMBackbone=MMaDA-8B-MixCoT, Search Width (K)=22025.12 | 57.54 | 8.18 | |
| ProbabilityBackbone=MMaDA-8B-MixCoT, Step Number (T)=2562025.12 | 56.18 | 11.26 | |
| MarginBackbone=MMaDA-8B-MixCoT, Step Number (T)=2562025.12 | 56.18 | 10.97 | |
| EntropyBackbone=MMaDA-8B-MixCoT, Step Number (T)=2562025.12 | 54.91 | 10.51 |