Debiasing Effectiveness on Out-of-Distribution (OOD) Split
29.31Mean RatioORIGINAL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ORIGINALModel Family=QWEN2.5-7B2026.03 | 29.31 | 81.74 | |
| ORIGINALModel Family=QWEN2.5-3B2026.03 | 16.06 | 37.6 | |
| ORIGINALModel Family=QWEN2.5-14B2026.03 | 13.5 | 37.52 | |
| ORIGINALModel Family=GEMMA-2-2B2026.03 | 4.48 | 8.38 | |
| UGIDModel Family=QWEN2.5-7B2026.03 | 2.14 | 1.42 | |
| KLAADModel Family=QWEN2.5-14B2026.03 | 1.61 | 3.96 | |
| CDAModel Family=QWEN2.5-3B2026.03 | 1.26 | 1.46 | |
| CDAModel Family=GEMMA-2-2B2026.03 | 1.18 | 1.21 | |
| KLAADModel Family=GEMMA-2-2B2026.03 | 1.18 | 1.29 | |
| CDAModel Family=QWEN2.5-14B2026.03 | 1.16 | 1.29 | |
| UGIDModel Family=QWEN2.5-3B2026.03 | 1.14 | 1.46 | |
| CDAModel Family=QWEN2.5-7B2026.03 | 1.14 | 1.29 | |
| KLAADModel Family=QWEN2.5-3B2026.03 | 1.04 | 1.87 | |
| UGIDModel Family=GEMMA-2-2B2026.03 | 1.03 | 1.13 | |
| KLAADModel Family=QWEN2.5-7B2026.03 | 1.01 | 1.29 | |
| UGIDModel Family=QWEN2.5-14B2026.03 | 1.01 | 1.07 |