Mathematical Reasoning on MATH (test) (TPF, TPS, Speedup, Solve Rate)
3.98Tokens Per Frame (TPF)Jacobi Forcing Model (MR)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Jacobi Forcing Model (MR)Family=AR-based, Backbone=Qwen2.5-Math-7B-Instruct, Optimization=Multi-block Decoding with Rejection recycling2025.12 | 3.98 | 152 | 3.68 | 77.4 | |
| Jacobi Forcing ModelFamily=AR-based, Backbone=Qwen2.5-Math-7B-Instruct2025.12 | 3.82 | 150.7 | 3.65 | 77.4 | |
| D2FFamily=Diffusion-based2025.12 | 2.6 | 98.8 | 2.39 | 35.4 | |
| CLLM*Family=AR-based, Backbone=Qwen2.5-Math-7B-Instruct2025.12 | 2.23 | 84.4 | 2.04 | 77.2 | |
| Fast-dLLM (DC)Family=Diffusion-based2025.12 | 1.9 | 67 | 1.62 | 37.1 | |
| JacobiFamily=AR-based, Backbone=Qwen2.5-Math-7B-Instruct2025.12 | 1.02 | 41 | 0.99 | 77 | |
| ARFamily=AR-based, Backbone=Qwen2.5-Math-7B-Instruct2025.12 | 1 | 41.3 | 1 | 77 | |
| LLaDA-InstructFamily=Diffusion-based2025.12 | 1 | 21.1 | 0.51 | 23.7 | |
| Dream-BaseFamily=Diffusion-based2025.12 | 1 | 9.9 | 0.24 | 35.8 |