Abstention in Question Answering on BB Answer Unknown
97.9Abstention F1RFT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| RFTModel=Llama-3.1-8B-Instruct2026.05 | 97.9 | 100 | 95.8 | 87 | |
| TIARModel=Llama-3.1-8B-Instruct2026.05 | 97.9 | 100 | 95.8 | 87 | |
| TruthRLModel=Llama-3.1-8B-Instruct2026.05 | 95.8 | 100 | 92 | 87 | |
| DPOModel=Llama-3.1-8B-Instruct2026.05 | 93.9 | 100 | 88.5 | 69.6 | |
| R-TuningModel=Llama-3.1-8B-Instruct2026.05 | 91.7 | 95.7 | 88 | 91.3 | |
| RFTModel=Qwen3-8B2026.05 | 89.8 | 95.7 | 84.6 | 78.3 | |
| TruthRLModel=Qwen3-8B2026.05 | 85.2 | 100 | 74.2 | 100 | |
| DPOModel=Qwen3-8B2026.05 | 83.6 | 100 | 71.9 | 95.7 | |
| R-TuningModel=Qwen3-8B2026.05 | 82.1 | 100 | 69.7 | 100 | |
| TIARModel=Qwen3-8B2026.05 | 79.3 | 100 | 65.7 | 100 |