Fact Verification on VitaminC
92.62AccuracyUD+-XXL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| UD+-XXLMode=Fully-supervised, Backbone=T5-XXL encoder, Model Parameters=11B2022.11 | 92.62 | — | |
| ALIGNModel Variant=large, Model Parameters=355M2023.07 | 91.8 | — | |
| Tay et al. (2022)Mode=Fully-supervised, Model Parameters=4x of UD (~44B)2022.11 | 91.1 | — | |
| ALIGNModel Variant=base, Model Parameters=125M2023.07 | 89.8 | — | |
| FLAN-T5Model Variant=xlarge, Model Parameters=3B2023.07 | 73.9 | — | |
| FLAN-T5Model Variant=large, Model Parameters=780M2023.07 | 72.9 | — | |
| LLaVA+LIMModality=Text-only, Augmentation=Latent Imagination Module2026.04 | 60 | 0.1016 | |
| BaselineModality=Text-only2026.04 | 51 | 0.4534 |