Abstention in Question Answering on QAQA False Premise
79.9Abstention F1R-Tuning
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| R-TuningModel=Qwen3-8B2026.05 | 79.9 | 90.2 | 71.8 | 82.5 | |
| TIARModel=Qwen3-8B2026.05 | 79.7 | 89.5 | 71.8 | 81.8 | |
| TruthRLModel=Qwen3-8B2026.05 | 79.1 | 87.7 | 72 | 84.6 | |
| DPOModel=Qwen3-8B2026.05 | 78 | 88.8 | 69.5 | 85.3 | |
| RFTModel=Qwen3-8B2026.05 | 70.3 | 70.5 | 70 | 67.7 | |
| TIARModel=Llama-3.1-8B-Instruct2026.05 | 58.2 | 46.7 | 77.3 | 48.4 | |
| DPOModel=Llama-3.1-8B-Instruct2026.05 | 58 | 47 | 75.7 | 46.7 | |
| TruthRLModel=Llama-3.1-8B-Instruct2026.05 | 58 | 46.7 | 76.4 | 50.5 | |
| RFTModel=Llama-3.1-8B-Instruct2026.05 | 57.1 | 45.6 | 76.5 | 44.9 | |
| R-TuningModel=Llama-3.1-8B-Instruct2026.05 | 56 | 45.6 | 72.6 | 44.2 |