Preference Bias Mitigation on Preference Bias
85.83RRSAC Single-task
Evaluation Results
| Method | Links | |
|---|---|---|
| SAC Single-taskBackbone=Qwen2-72B-Instruct2024.11 | 85.83 | |
| SAC Multi-taskBackbone=Qwen2-72B-Instruct2024.11 | 80.83 | |
| SACControl Dim=Single2024.11 | 62.5 | |
| SAC Multi-taskBackbone=Qwen2-7B-Instruct2024.11 | 58.33 | |
| SACControl Dim=Multiple2024.11 | 53.75 | |
| SAC Single-taskBackbone=Qwen2-7B-Instruct2024.11 | 50.83 | |
| RepEControl Dim=Single2024.11 | 39.17 | |
| No ControlControl Dim=Single2024.11 | 10.83 | |
| No ControlControl Dim=Multiple2024.11 | 10.83 | |
| RepE-MergeControl Dim=Multiple2024.11 | 10 | |
| RepE-MeanControl Dim=Multiple2024.11 | 5 | |
| No ControlBackbone=Qwen2-7B-Instruct2024.11 | 0 | |
| No ControlBackbone=Qwen2-72B-Instruct2024.11 | 0 |