Veracity Prediction on MOCHEG
57Macro F1 ScoreLlama-8BQwenVL
Evaluation Results
| Method | Links | |
|---|---|---|
| Llama-8BQwenVLBackbone=Llama-8B, Justification Generator=QwenVL, Protocol=QLoRA2025.05 | 57 | |
| Mistral-7BPaligemmaBackbone=Mistral-7B, Justification Generator=Paligemma, Protocol=QLoRA2025.05 | 57 | |
| Llama-8BPaligemmaBackbone=Llama-8B, Justification Generator=Paligemma, Protocol=QLoRA2025.05 | 56 | |
| Mistral-7BQwenVLBackbone=Mistral-7B, Justification Generator=QwenVL, Protocol=QLoRA2025.05 | 56 | |
| Mistral-7BIdefics3Backbone=Mistral-7B, Justification Generator=Idefics3, Protocol=QLoRA2025.05 | 56 | |
| Llama-8BIdefics3Backbone=Llama-8B, Justification Generator=Idefics3, Protocol=QLoRA2025.05 | 55 | |
| RoBERTa-LIdefics3Backbone=RoBERTa-Large, Justification Generator=Idefics3, Protocol=Fine-tuning2025.05 | 51 | |
| XLNet-LIdefics3Backbone=XLNet-Large, Justification Generator=Idefics3, Protocol=Fine-tuning2025.05 | 50 | |
| RoBERTa-LPaligemmaBackbone=RoBERTa-Large, Justification Generator=Paligemma, Protocol=Fine-tuning2025.05 | 48 | |
| XLNet-LQwenVLBackbone=XLNet-Large, Justification Generator=QwenVL, Protocol=Fine-tuning2025.05 | 48 | |
| XLNet-LPaligemmaBackbone=XLNet-Large, Justification Generator=Paligemma, Protocol=Fine-tuning2025.05 | 48 | |
| RoBERTa-LQwenVLBackbone=RoBERTa-Large, Justification Generator=QwenVL, Protocol=Fine-tuning2025.05 | 47 | |
| Cekinel et al., 2025Protocol=Baseline2025.05 | 37 |