Mathematical Reasoning on MATH 500 (Accuracy, Latency, Tokens, ActRatio)
61.6AccuracyOriginal
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OriginalBackbone=Qwen2.5-7B Series, Ratio=-2025.10 | 61.6 | 571.64 | 6.351 | — | |
| PromptBackbone=Qwen2.5-7B Series, Ratio=0.92025.10 | 60.401 | 553.27 | 6.271 | 97 | |
| CoDBackbone=Qwen2.5-7B Series, Ratio=0.52025.10 | 51.301 | 314.4 | 3.611 | 55 | |
| TruncationBackbone=Qwen2.5-7B Series, Ratio=0.92025.10 | 42.7018 | 514.48 | 5.721 | 90 | |
| UCoTBackbone=Llama3.1-8B Series, Ratio=0.52025.10 | 38.004 | 298.63 | 3.791 | 52 |