LLM Alignment on Capybara
28.34Win RateClean Data
Evaluation Results
| Method | Links | |
|---|---|---|
| Clean Data2026.06 | 28.34 | |
| DeMix-unifrepair=true2026.06 | 27.95 | |
| DeMixrepair=true2026.06 | 27.53 | |
| DeMix-unifrepair=false2026.06 | 26.82 | |
| DeMixrepair=false2026.06 | 26.51 | |
| DDA-select2026.06 | 25.53 | |
| Error. Data2026.06 | 25.52 | |
| DDA-repair2026.06 | 24.57 |