Mathematical Verification on MATH (test)
79.9Mean Hit RateCost-Stratified Thresholding
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Cost-Stratified ThresholdingModel=LLaMA3-8B, beta=10%, err=.613, rho=.2012026.06 | 79.9 | 4.9 | |
| Cost-Stratified ThresholdingModel=LLaMA3-8B, beta=20%, err=.613, rho=.2012026.06 | 75.8 | 1.6 | |
| Cost-Stratified ThresholdingModel=LLaMA3-8B, beta=30%, err=.613, rho=.2012026.06 | 73.4 | 1.5 | |
| Cost-Stratified ThresholdingModel=LLaMA3-8B, beta=50%, err=.613, rho=.2012026.06 | 69.3 | 3.7 | |
| Cost-Stratified ThresholdingModel=Qwen3-8B, beta=10%, err=.247, rho=.1532026.06 | 32.2 | 20.2 | |
| Cost-Stratified ThresholdingModel=Qwen3-8B, beta=20%, err=.247, rho=.1532026.06 | 30.8 | 7.6 | |
| Cost-Stratified ThresholdingModel=Qwen3-8B, beta=30%, err=.247, rho=.1532026.06 | 30.1 | 1.9 | |
| Cost-Stratified ThresholdingModel=Qwen3-8B, beta=50%, err=.247, rho=.1532026.06 | 28.3 | 2 |