Reasoning on MMLU-Pro (ER, TP %)
3ERB-PAC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| B-PACRisk tolerance (epsilon)=0.08, Confidence level (alpha)=0.12026.01 | 3 | 77.24 | |
| CoDRisk tolerance (epsilon)=0.08, Confidence level (alpha)=0.12026.01 | 10 | 73.63 | |
| NoThinkingRisk tolerance (epsilon)=0.08, Confidence level (alpha)=0.12026.01 | 12 | 76.52 |