Logical Reasoning on SATA-Bench
25.2AccuracySFT+RL
Evaluation Results
| Method | Links | |
|---|---|---|
| SFT+RLType=Fast CoT, Instant Response=√2026.05 | 25.2 | |
| SFT+RLType=Thinking, Instant Response=×2026.05 | 24 | |
| InterRSType=Interleaved, Instant Response=√2026.05 | 21.1 | |
| SFTType=Interleaved, Instant Response=√2026.05 | 17.5 | |
| Zero-shotType=Thinking, Instant Response=×2026.05 | 0.3 | |
| Zero-shotType=Direct, Instant Response=√2026.05 | 0.2 | |
| Zero-shotType=Fast CoT, Instant Response=√2026.05 | 0.1 |