Multi-turn Calibration on SciQ
4.42ECE@1MTCal
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MTCalBackbone=Gemma2-9B-it2026.04 | 4.42 | 2.29 | 3.84 | 12.17 | 3.56 | |
| DCalBackbone=Gemma2-9B-it2026.04 | 4.68 | 12.72 | 5.33 | 12.79 | 5.4 | |
| MTCalBackbone=Qwen2.5-7B-Instruct2026.04 | 4.78 | 5.29 | 5.68 | 19.97 | 4.46 | |
| MTCalBackbone=Llama3.1-8B-Instruct2026.04 | 5.39 | 6.39 | 5.65 | 20.83 | 5.43 | |
| DCalBackbone=Qwen2.5-7B-Instruct2026.04 | 5.46 | 14.52 | 5.7 | 20.72 | 5.58 | |
| DCalBackbone=Llama3.1-8B-Instruct2026.04 | 5.8 | 16.3 | 6.23 | 21.81 | 6.2 | |
| SLBackbone=Llama3.1-8B-Instruct2026.04 | 7.32 | 21.56 | 12.05 | 23.43 | 10.51 | |
| PSBackbone=Qwen2.5-7B-Instruct2026.04 | 7.89 | 10.26 | 8.03 | 20.04 | 7.19 | |
| PSBackbone=Llama3.1-8B-Instruct2026.04 | 10.03 | 13.36 | 12.5 | 21.47 | 12.67 | |
| SLBackbone=Qwen2.5-7B-Instruct2026.04 | 12.09 | 24.16 | 15.02 | 22.62 | 11.18 | |
| P(True)Backbone=Gemma2-9B-it2026.04 | 14.66 | 19.36 | 13.89 | 14.64 | 12.4 | |
| VerbalBackbone=Gemma2-9B-it2026.04 | 14.96 | 19.83 | 16.39 | 10.73 | 8.65 | |
| SLBackbone=Gemma2-9B-it2026.04 | 15.02 | 21.08 | 18.74 | 16.64 | 11.91 | |
| VerbalBackbone=Llama3.1-8B-Instruct2026.04 | 15.12 | 32.09 | 20.3 | 23.35 | 19.69 | |
| PSBackbone=Gemma2-9B-it2026.04 | 16.75 | 18.81 | 17.4 | 15.43 | 12.87 | |
| P(True)Backbone=Qwen2.5-7B-Instruct2026.04 | 17.94 | 21.99 | 17.12 | 20.53 | 16.21 | |
| P(True)Backbone=Llama3.1-8B-Instruct2026.04 | 18.19 | 22.37 | 17.91 | 21.27 | 17.17 | |
| VerbalBackbone=Qwen2.5-7B-Instruct2026.04 | 20.16 | 35.48 | 24.36 | 23.63 | 21.29 | |
| SCBackbone=Qwen2.5-7B-Instruct2026.04 | 21.37 | 24.2 | 23.52 | 27.8 | 18.74 | |
| SCBackbone=Llama3.1-8B-Instruct2026.04 | 32.98 | 30.1 | 38.71 | 39.1 | 29.38 | |
| SCBackbone=Gemma2-9B-it2026.04 | 44.27 | 47.41 | 55.56 | 47.81 | 38.97 |