Fairness Evaluation on UnQover
99.6ScoreSelf-Debias Offline + Self-Correction
Evaluation Results
| Method | Links | |
|---|---|---|
| Self-Debias Offline + Self-CorrectionOptimization Stage=Offline, Self-Correction=True2026.04 | 99.6 | |
| Self-Debias Iter1Optimization Stage=Iter1, Self-Correction=False2026.04 | 99.6 | |
| Self-Debias SFTOptimization Stage=SFT, Self-Correction=False2026.04 | 99.5 | |
| Self-Debias SFT + Self-CorrectionOptimization Stage=SFT, Self-Correction=True2026.04 | 99.5 | |
| Self-Debias OfflineOptimization Stage=Offline, Self-Correction=False2026.04 | 99.5 | |
| Self-Debias Iter1 + Self-CorrectionOptimization Stage=Iter1, Self-Correction=True2026.04 | 99.5 | |
| Self-Debias Iter2Optimization Stage=Iter2, Self-Correction=False2026.04 | 99.5 | |
| Self-Debias Iter2 + Self-CorrectionOptimization Stage=Iter2, Self-Correction=True2026.04 | 99.5 | |
| Qwen1.5-8BSelf-Correction=False2026.04 | 97.3 | |
| Qwen2.5-7B-Instruct + Self-CorrectionSelf-Correction=True2026.04 | 97 | |
| Qwen1.5-8B + Self-CorrectionSelf-Correction=True2026.04 | 95.4 | |
| Qwen2.5-7B-InstructSelf-Correction=False2026.04 | 93.9 | |
| DeepSeek-R1-Distill-Qwen-7BSelf-Correction=False2026.04 | 83.9 | |
| DeepSeek-R1-Distill-Qwen-7B + Self-CorrectionSelf-Correction=True2026.04 | 82.2 | |
| FairGPTQ-INT4Model=Qwen-3-8B2025.09 | 77.76 | |
| FairGPTQ-INT4Model=Qwen-2.5-7B-Instruct2025.09 | 73.55 | |
| Base-FP16Model=Qwen-2.5-7B-Instruct2025.09 | 73.14 | |
| GPTQ-INT4Model=Qwen-2.5-7B-Instruct2025.09 | 71.71 | |
| GPTQ-INT4Model=Qwen-3-8B2025.09 | 70.06 | |
| Base-FP16Model=Qwen-3-8B2025.09 | 66.7 | |
| FairGPTQ-INT4Model=Mistral-v0.3-7B-Instruct2025.09 | 58.89 | |
| Llama-3.1-8B-Instruct + Self-CorrectionSelf-Correction=True2026.04 | 57.8 | |
| FairGPTQ-INT4Model=OPT-6.7B2025.09 | 36.27 | |
| Base-FP16Model=Mistral-v0.3-7B-Instruct2025.09 | 35.58 | |
| Base-FP16Model=OPT-6.7B2025.09 | 34.63 | |
| Llama-3.1-8B-InstructSelf-Correction=False2026.04 | 33.5 | |
| GPTQ-INT4Model=Mistral-v0.3-7B-Instruct2025.09 | 33.3 | |
| GPTQ-INT4Model=OPT-6.7B2025.09 | 32.1 | |
| Base-FP16Model=LLaMA-3.1-8B-Instruct2025.09 | 30.06 | |
| GPTQ-INT4Model=Mistral-v0.3-7B2025.09 | 27.61 | |
| FairGPTQ-INT4Model=Mistral-v0.3-7B2025.09 | 26.7 | |
| FairGPTQ-INT4Model=LLaMA-3.1-8B-Instruct2025.09 | 26.66 | |
| Base-FP16Model=Mistral-v0.3-7B2025.09 | 24.85 | |
| GPTQ-INT4Model=LLaMA-3.1-8B-Instruct2025.09 | 14.03 |