Logical Reasoning on RP LP Domain, Npred ≤ 60, δ ≤ 6
99.4Accuracycot w. L = 64
Evaluation Results
| Method | Links | |
|---|---|---|
| cot w. L = 64Reasoning Strategy=CoT, Number of Layers=642026.05 | 99.4 | |
| direct w. L = 128Reasoning Strategy=Direct, Number of Layers=1282026.05 | 99.2 | |
| direct w. L = 128Reasoning Strategy=Direct, Number of Layers=1282026.05 | 98.9 | |
| cot w. L = 128Reasoning Strategy=CoT, Number of Layers=1282026.05 | 98.6 | |
| direct w. L = 64Reasoning Strategy=Direct, Number of Layers=642026.05 | 98.5 | |
| direct w. L = 96Reasoning Strategy=Direct, Number of Layers=962026.05 | 98.4 | |
| cot w. L = 64Reasoning Strategy=CoT, Number of Layers=642026.05 | 97.8 | |
| direct w. L = 48Reasoning Strategy=Direct, Number of Layers=482026.05 | 97.6 | |
| direct w. L = 32Reasoning Strategy=Direct, Number of Layers=322026.05 | 97.5 | |
| cot w. L = 128Reasoning Strategy=CoT, Number of Layers=1282026.05 | 97 | |
| direct w. L = 24Reasoning Strategy=Direct, Number of Layers=242026.05 | 96.7 | |
| cot w. L = 96Reasoning Strategy=CoT, Number of Layers=962026.05 | 96.7 | |
| cot w. L = 48Reasoning Strategy=CoT, Number of Layers=482026.05 | 96.6 | |
| cot w. L = 24Reasoning Strategy=CoT, Number of Layers=242026.05 | 95.7 | |
| cot w. L = 32Reasoning Strategy=CoT, Number of Layers=322026.05 | 95.4 | |
| cot (baseline)Reasoning Strategy=CoT2026.05 | 94.7 | |
| cot w. L = 16Reasoning Strategy=CoT, Number of Layers=162026.05 | 92.4 | |
| direct w. L = 16Reasoning Strategy=Direct, Number of Layers=162026.05 | 92.2 | |
| direct (baseline)Reasoning Strategy=Direct2026.05 | 91.8 | |
| cot (baseline)Reasoning Strategy=CoT2026.05 | 88.2 | |
| cot w. Nepochs = 60Reasoning Strategy=CoT, N_epochs=602026.05 | 85.3 | |
| cot w. Nepochs = 30Reasoning Strategy=CoT, N_epochs=302026.05 | 85.1 | |
| direct (baseline)Reasoning Strategy=Direct2026.05 | 80.9 | |
| direct w. Nepochs = 30Reasoning Strategy=Direct, N_epochs=302026.05 | 78.3 | |
| direct w. Nepochs = 60Reasoning Strategy=Direct, N_epochs=602026.05 | 77.3 |