List your company
ResearchFormerly WizWand

Mathematical Reasoning on GSM8K (Throughput, Latency, Score)

12.37Throughputd2Cache
1.90764.62387.3410.0562Sep 27, 2025
Updated 8mo ago

Evaluation results

MethodLinks
d2CacheBackbone=Dream-Base, Few-shot=4-shot, Generation Length=2562025.0912.3721.7473.5
d2CacheBackbone=LLaDA-Base, Few-shot=4-shot, Generation Length=2562025.0911.3722.5772.1
Fast-dLLMBackbone=Dream-Base, Few-shot=4-shot, Generation Length=2562025.0910.0725.4471.4
Fast-dLLMBackbone=LLaDA-Base, Few-shot=4-shot, Generation Length=2562025.099.9425.8270.3
dLLM-CacheBackbone=Dream-Base, Few-shot=4-shot, Generation Length=2562025.099.2827.8864.7
dLLM-CacheBackbone=LLaDA-Base, Few-shot=4-shot, Generation Length=2562025.097.7233.369.3
VanillaBackbone=Dream-Base, Few-shot=4-shot, Generation Length=2562025.092.6796.2971.7
VanillaBackbone=LLaDA-Base, Few-shot=4-shot, Generation Length=2562025.092.31112.3970.4