Mathematical Reasoning on GSM8K (4-shot, max_gen_len=256)
79.15Accuracy (%)Elastic-Cache
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Elastic-CacheModel=LLaDA-8B-Instruct2026.03 | 79.15 | 23.49 | 10.3 | |
| EntropyCacheModel=LLaDA-8B-Instruct2026.03 | 78.77 | 38.49 | 16.9 | |
| Baseline+ParallelModel=LLaDA-8B-Instruct2026.03 | 78.47 | 8.56 | 3.8 | |
| Fast-DLLM (Dual)Model=LLaDA-8B-Instruct2026.03 | 78.01 | 32.54 | 14.3 | |
| d2CacheModel=LLaDA-8B-Instruct2026.03 | 78.01 | 29.96 | 13.2 | |
| BaselineModel=LLaDA-8B-Instruct2026.03 | 77.86 | 2.28 | — | |
| d2CacheModel=Dream-v0-Instruct-7B2026.03 | 75.44 | 41.68 | 16 | |
| Baseline+ParallelModel=Dream-v0-Instruct-7B2026.03 | 74.91 | 12.14 | 4.7 | |
| BaselineModel=Dream-v0-Instruct-7B2026.03 | 74.3 | 2.6 | — | |
| EntropyCacheModel=Dream-v0-Instruct-7B2026.03 | 74.3 | 48.32 | 18.6 | |
| Fast-DLLM (Dual)Model=Dream-v0-Instruct-7B2026.03 | 72.25 | 42.17 | 16.2 | |
| Elastic-CacheModel=Dream-v0-Instruct-7B2026.03 | 69.6 | 32.15 | 12.4 |