Calibration on Average (StrategyQA, HotpotQA, NQ, Bamboogle)
0.264ECENAACL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| NAACLBackbone=Qwen2.5-7B-Instruct2026.01 | 0.264 | 0.768 | |
| NAACLBackbone=Llama-3.1-8B-Instruct2026.01 | 0.266 | 0.751 | |
| NAACLBackbone=DeepSeek-R1-Distill-Llama-8B2026.01 | 0.311 | 0.679 | |
| Noise-awareBackbone=Llama-3.1-8B-Instruct2026.01 | 0.314 | 0.667 | |
| Label-only SFTBackbone=Qwen2.5-7B-Instruct2026.01 | 0.315 | 0.722 | |
| Noise-awareBackbone=Qwen2.5-7B-Instruct2026.01 | 0.323 | 0.697 | |
| CoTBackbone=Qwen2.5-7B-Instruct2026.01 | 0.335 | 0.726 | |
| NAACLBackbone=DeepSeek-R1-Distill-Qwen-7B2026.01 | 0.344 | 0.723 | |
| EnsembleBackbone=Llama-3.1-8B-Instruct2026.01 | 0.352 | 0.648 | |
| Label-only SFTBackbone=Llama-3.1-8B-Instruct2026.01 | 0.353 | 0.686 | |
| EnsembleBackbone=Qwen2.5-7B-Instruct2026.01 | 0.36 | 0.722 | |
| VanillaBackbone=Qwen2.5-7B-Instruct2026.01 | 0.366 | 0.73 | |
| Noise-awareBackbone=DeepSeek-R1-Distill-Llama-8B2026.01 | 0.373 | 0.615 | |
| CoTBackbone=Llama-3.1-8B-Instruct2026.01 | 0.377 | 0.591 | |
| Noise-awareBackbone=DeepSeek-R1-Distill-Qwen-7B2026.01 | 0.388 | 0.571 | |
| VanillaBackbone=DeepSeek-R1-Distill-Llama-8B2026.01 | 0.407 | 0.657 | |
| EnsembleBackbone=DeepSeek-R1-Distill-Llama-8B2026.01 | 0.407 | 0.648 | |
| VanillaBackbone=Llama-3.1-8B-Instruct2026.01 | 0.411 | 0.605 | |
| CoTBackbone=DeepSeek-R1-Distill-Llama-8B2026.01 | 0.436 | 0.65 | |
| VanillaBackbone=DeepSeek-R1-Distill-Qwen-7B2026.01 | 0.437 | 0.654 | |
| Label-only SFTBackbone=DeepSeek-R1-Distill-Llama-8B2026.01 | 0.44 | 0.603 | |
| EnsembleBackbone=DeepSeek-R1-Distill-Qwen-7B2026.01 | 0.462 | 0.623 | |
| CoTBackbone=DeepSeek-R1-Distill-Qwen-7B2026.01 | 0.474 | 0.655 | |
| Label-only SFTBackbone=DeepSeek-R1-Distill-Qwen-7B2026.01 | 0.517 | 0.573 |