Human Evaluation on NeurIPS ICML ICLR Proposals 2025 (test)
31WinsStepwise CoT
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Stepwise CoTComparison=Stepwise CoT vs. Prompting, Dimension=Overall2026.03 | 31 | 3 | 26 | 54.2 | 26.7 | |
| Stepwise CoTComparison=Stepwise CoT vs. Prompting, Dimension=Excitement2026.03 | 30 | 10 | 20 | 58.3 | 26.7 | |
| Stepwise CoTComparison=Stepwise CoT vs. Prompting, Dimension=Soundness2026.03 | 29 | 7 | 24 | 54.2 | 31.7 | |
| Stepwise CoTComparison=Stepwise CoT vs. Human, Dimension=Overall2026.03 | 25 | 10 | 25 | 50 | 11.7 | |
| Stepwise CoTComparison=Stepwise CoT vs. Human, Dimension=Soundness2026.03 | 21 | 11 | 28 | 44.2 | 11.7 | |
| Stepwise CoTComparison=Stepwise CoT vs. Human, Dimension=Excitement2026.03 | 20 | 18 | 22 | 48.3 | 6.7 |