Mathematical Reasoning on GSM8K (CoT)
99.49AccuracyAverage
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| AverageBackbone=LLaMA-32026.03 | 99.49 | — | — | |
| Task ArithmeticBackbone=LLaMA-32026.03 | 98.77 | — | — | |
| ACEBackbone=LLaMA-32026.03 | 94.26 | — | — | |
| DreamBackbone=Dream2026.05 | 83.9 | 1 | 83.9 | |
| dParallelBackbone=Dream2026.05 | 82.1 | 3.02 | 245.7 | |
| d3LLMBackbone=Dream2026.05 | 81.4 | 4.94 | 391.3 | |
| TAD-QBackbone=Dream, Configuration=Quality2026.05 | 81.4 | 3.26 | 255.6 | |
| TAD-SBackbone=Dream, Configuration=Speed2026.05 | 81 | 4.59 | 359.5 | |
| TAD-QBackbone=LLaDA, Configuration=Quality2026.05 | 79.9 | 6.32 | 503.8 | |
| Fast-dLLMBackbone=Dream2026.05 | 79 | 1.44 | 116.5 | |
| TAD-SBackbone=LLaDA, Configuration=Speed2026.05 | 78.8 | 8.47 | 649 | |
| Fast-dLLM-v2Backbone=Dream2026.05 | 77.5 | 2.21 | 156 | |
| Fast-dLLMBackbone=LLaDA2026.05 | 74.7 | 2.77 | 182.6 | |
| mathModel Type=Single-task fine-tuning expert, Backbone=LLaMA-32026.03 | 74 | — | — | |
| AverageModel Type=Merged model, Backbone=LLaMA-32026.03 | 73.62 | — | — | |
| D2FBackbone=LLaDA2026.05 | 73.2 | 2.88 | 189.2 | |
| d3LLMBackbone=LLaDA2026.05 | 73.1 | 9.11 | 539.2 | |
| Task ArithmeticModel Type=Merged model, Backbone=LLaMA-32026.03 | 73.09 | — | — | |
| LLaDABackbone=LLaDA-Instruct2026.05 | 72.6 | 1 | 72.6 | |
| dParallelBackbone=LLaDA2026.05 | 72.6 | 5.14 | 308.7 | |
| ACEModel Type=Merged model, Backbone=LLaMA-32026.03 | 69.75 | — | — |