Truthful Question Answering on TruthfulQA (accuracy, intervention rate)
70AccuracyAligned base model
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Aligned base modelModel Pair=Q→G, Pair Family=Cross-Family2026.04 | 70 | — | |
| Aligned base modelModel Pair=L→G, Pair Family=Cross-Family2026.04 | 70 | — | |
| Aligned base modelModel Pair=G→G, Pair Family=Within-Family2026.04 | 70 | — | |
| Aligned base modelModel Pair=Q→L, Pair Family=Cross-Family2026.04 | 67 | — | |
| Aligned base modelModel Pair=G→L, Pair Family=Cross-Family2026.04 | 67 | — | |
| Aligned base modelModel Pair=L→L, Pair Family=Within-Family2026.04 | 67 | — | |
| Base ModelModel Pair=Q→L, Pair Family=Cross-Family2026.04 | 58 | — | |
| Base ModelModel Pair=G→L, Pair Family=Cross-Family2026.04 | 58 | — | |
| Base ModelModel Pair=L→L, Pair Family=Within-Family2026.04 | 58 | — | |
| NUDGINGModel Pair=L→L, Pair Family=Within-Family2026.04 | 51 | 20 | |
| BlendInModel Pair=L→L, Pair Family=Within-Family2026.04 | 51 | 20 | |
| BlendInModel Pair=Q→L, Pair Family=Cross-Family2026.04 | 50 | 31.9 | |
| BlendInModel Pair=G→L, Pair Family=Cross-Family2026.04 | 50 | 19.2 | |
| Guidance ModelModel Pair=L→G, Pair Family=Cross-Family2026.04 | 50 | — | |
| Guidance ModelModel Pair=L→L, Pair Family=Within-Family2026.04 | 50 | — | |
| BlendInModel Pair=L→G, Pair Family=Cross-Family2026.04 | 49 | 17.2 | |
| Guidance ModelModel Pair=Q→L, Pair Family=Cross-Family2026.04 | 48 | — | |
| NUDGINGModel Pair=Q→L, Pair Family=Cross-Family2026.04 | 48 | 31.9 | |
| Guidance ModelModel Pair=Q→G, Pair Family=Cross-Family2026.04 | 48 | — | |
| NUDGINGModel Pair=L→G, Pair Family=Cross-Family2026.04 | 47 | 17.2 | |
| Guidance ModelModel Pair=G→L, Pair Family=Cross-Family2026.04 | 46 | — | |
| Guidance ModelModel Pair=G→G, Pair Family=Within-Family2026.04 | 46 | — | |
| NUDGINGModel Pair=G→L, Pair Family=Cross-Family2026.04 | 45 | 19.2 | |
| NUDGINGModel Pair=G→G, Pair Family=Within-Family2026.04 | 43 | 20.9 | |
| BlendInModel Pair=G→G, Pair Family=Within-Family2026.04 | 43 | 20.9 | |
| BlendInModel Pair=Q→G, Pair Family=Cross-Family2026.04 | 42 | 31.5 | |
| Base ModelModel Pair=Q→G, Pair Family=Cross-Family2026.04 | 40 | — | |
| NUDGINGModel Pair=Q→G, Pair Family=Cross-Family2026.04 | 40 | 31.5 | |
| Base ModelModel Pair=L→G, Pair Family=Cross-Family2026.04 | 40 | — | |
| Base ModelModel Pair=G→G, Pair Family=Within-Family2026.04 | 40 | — |