List your company

Mathematical Reasoning on GSM8K (Accuracy, Throughput)

95.53AccuracyQwen2.5 7B
35.719651.247366.77582.3027Jun 6, 2026

Evaluation results

MethodLinks
Qwen2.5 7BGeneration mode=Autoregressive2026.0695.5349.26
PoE-BridgeImportance sampling=true2026.0695.3100.71
PoE-BridgeImportance sampling=false2026.0695.2104.49
Dream 7BSampling strategy=Entropy, Tokens per step=22026.067226.25
Dream 7BSampling strategy=Entropy, Tokens per step=42026.0652.2747.37
Dream 7BSampling strategy=Random, Tokens per step=22026.0638.0215.42