Adversarial Factuality
99.04CRSAC Single-task
Evaluation Results
| Method | Links | |
|---|---|---|
| SAC Single-taskBackbone=Qwen2-72B-Instruct2024.11 | 99.04 | |
| SAC Multi-taskBackbone=Qwen2-72B-Instruct2024.11 | 98.56 | |
| No ControlBackbone=Qwen2-72B-Instruct2024.11 | 98.08 | |
| SAC Multi-taskBackbone=Qwen2-7B-Instruct2024.11 | 96.65 | |
| SAC Single-taskBackbone=Qwen2-7B-Instruct2024.11 | 95.22 | |
| RepEControl Dim=Single2024.11 | 90.43 | |
| SACControl Dim=Single2024.11 | 89.47 | |
| SACControl Dim=Multiple2024.11 | 86.12 | |
| No ControlControl Dim=Single2024.11 | 76.56 | |
| No ControlControl Dim=Multiple2024.11 | 76.56 | |
| No ControlBackbone=Qwen2-7B-Instruct2024.11 | 76.08 | |
| RepE-MeanControl Dim=Multiple2024.11 | 72.59 | |
| RepE-MergeControl Dim=Multiple2024.11 | 71.08 |