Factual Prediction on KLAR (all samples)
76.08Accuracyshortcut method
Evaluation results
| Method | Links | |
|---|---|---|
| shortcut methodBackbone=LLaMA22025.04 | 76.08 | |
| fine-tuningBackbone=LLaMA2, tuning_type=LoRA2025.04 | 72.26 | |
| original modelBackbone=LLaMA22025.04 | 71.47 | |
| translation-enBackbone=LLaMA22025.04 | 53.88 | |
| shortcut methodBackbone=BLOOM2025.04 | 51.67 | |
| original modelBackbone=BLOOM2025.04 | 43.24 | |
| fine-tuningBackbone=BLOOM, tuning_type=full model2025.04 | 31.73 | |
| translation-enBackbone=BLOOM2025.04 | 28.03 | |
| translation-early-exitBackbone=BLOOM2025.04 | 15.68 | |
| translation-early-exitBackbone=LLaMA22025.04 | 13.93 |