Mathematical Reasoning on FINQA
72.2AccuracyCPMax
Evaluation results
| Method | Links | |
|---|---|---|
| CPMaxModel=Average across models, Sampling=Cross-modal maximization, Max Samples=402026.04 | 72.2 | |
| CPMajModel=Average across models, Sampling=Cross-modal majority voting, Max Samples=402026.04 | 72.1 | |
| CPAgrModel=Average across models, Sampling=Cross-modal agreement, Max Samples=402026.04 | 72 | |
| SCPoTModel=Average across models, Sampling=Self-Consistency PoT, Max Samples=402026.04 | 71.7 | |
| SCCoTModel=Average across models, Sampling=Self-Consistency CoT, Max Samples=402026.04 | 70.4 | |
| PoTModel=Average across models, Sampling=Single temperature-0 sample2026.04 | 70 | |
| CoTModel=Average across models, Sampling=Single temperature-0 sample2026.04 | 68 |