CoT Length Prediction on GPQA Diamond
27.32Relative MAEFuel Gauge
Evaluation Results
| Method | Links | |
|---|---|---|
| Fuel GaugeBackbone Model=Qwen3-8B2026.03 | 27.32 | |
| Fuel GaugeBackbone Model=Qwen3-4B2026.03 | 31.85 | |
| DirectBackbone Model=Qwen3-4B2026.03 | 49.32 | |
| MeanBackbone Model=Qwen3-8B2026.03 | 50.03 | |
| MedianBackbone Model=Qwen3-8B2026.03 | 52.12 | |
| MeanBackbone Model=Qwen3-4B2026.03 | 57.21 | |
| DirectBackbone Model=Qwen3-8B2026.03 | 57.95 | |
| MedianBackbone Model=Qwen3-4B2026.03 | 59.67 |