Mathematical Reasoning on Meta-audio
61.5AccuracySFT+RL
Evaluation results
| Method | Links | |
|---|---|---|
| SFT+RLType=Thinking, Instant Response=×2026.05 | 61.5 | |
| InterRSType=Interleaved, Instant Response=√2026.05 | 59.2 | |
| SFTType=Interleaved, Instant Response=√2026.05 | 59.1 | |
| Zero-shotType=Thinking, Instant Response=×2026.05 | 52.6 | |
| SFT+RLType=Fast CoT, Instant Response=√2026.05 | 40.7 | |
| Zero-shotType=Fast CoT, Instant Response=√2026.05 | 19.7 | |
| Zero-shotType=Direct, Instant Response=√2026.05 | 10.5 |