Multi-task Evaluation on Fairness and Utility Suite
82.1Average ScoreSelf-Debias Iter2 + Self-Correction
Evaluation Results
| Method | Links | |
|---|---|---|
| Self-Debias Iter2 + Self-Correction2026.04 | 82.1 | |
| Self-Debias Iter1 + Self-Correction2026.04 | 81.8 | |
| Self-Debias Iter22026.04 | 81.7 | |
| Self-Debias Offline + Self-Correction2026.04 | 81.3 | |
| Self-Debias Iter12026.04 | 81.3 | |
| Self-Debias SFT + Self-Correction2026.04 | 80.9 | |
| Self-Debias Offline2026.04 | 80.8 | |
| Self-Debias SFT2026.04 | 80.6 | |
| Qwen1.5-8B2026.04 | 77.5 | |
| Qwen2.5-7B-Instruct2026.04 | 77.4 | |
| Qwen2.5-7B-Instruct + Self-Correction2026.04 | 70.9 | |
| DeepSeek-R1-Distill-Qwen-7B2026.04 | 70.4 | |
| Qwen1.5-8B + Self-Correction2026.04 | 64 | |
| DeepSeek-R1-Distill-Qwen-7B + Self-Correction2026.04 | 63.7 | |
| Llama-3.1-8B-Instruct2026.04 | 52.3 | |
| Llama-3.1-8B-Instruct + Self-Correction2026.04 | 42.8 |