Veracity prediction on RAWFC
65.5Macro F1 ScoreG-Defensebackground
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| G-DefensebackgroundApproach category=Ours2026.04 | 65.5 | 66.29 | 65.49 | |
| S-EGSx -> y=c → v; exp, Model (Backbone)=LLaMA2-7B, No Closed-Source API Dependency=true, # Distill Explanations for RAW-FC=4652025.11 | 64.99 | 65.04 | 65.01 | |
| G-Defensedecomp+Approach category=Ours2026.04 | 64.34 | 65.63 | 64.33 | |
| G-DefenseApproach category=Ours2026.04 | 64.31 | 64.99 | 64.52 | |
| G-DefensehyperApproach category=Ours2026.04 | 63.44 | 64.62 | 63.02 | |
| L-DenfenseGPT-3.5Approach category=LLM-based approach2026.04 | 61.2 | 61.72 | 61.01 | |
| L-Denfensex -> y=c; evi → v; exp, Model (Backbone)=ChatGPT + Roberta-Large, No Closed-Source API Dependency=false, # Distill Explanations for RAW-FC=32,2402025.11 | 61.2 | 61.72 | 61.01 | |
| G-DefenseLLaMA3.1Approach category=Ours2026.04 | 60.76 | 60.97 | 61.23 | |
| SFTx -> y=c → v; exp, Model (Backbone)=LLaMA2-7B, No Closed-Source API Dependency=true, # Distill Explanations for RAW-FC=02025.11 | 60.59 | 60.66 | 61.04 | |
| L-DenfenseLLaMA2Approach category=LLM-based approach2026.04 | 60.12 | 60.95 | 60 | |
| RAVx -> y=c; evi → v; exp, Model (Backbone)=LLaMA-3.1-70B-Instruct x3, No Closed-Source API Dependency=true2025.11 | 59.19 | — | — | |
| FactLLaMAknowApproach category=LLM-based approach2026.04 | 55.65 | 56.11 | 55.5 | |
| FactLLaMA (Google)x -> y=c; evi → v, Model (Backbone)=LLaMA2-7B, No Closed-Source API Dependency=false, utilizing search API=Google2025.11 | 55.65 | 56.11 | 55.5 | |
| HiSS (Google)x -> y=c → v; exp, Model (Backbone)=ChatGPT, No Closed-Source API Dependency=false, utilizing search API=Google2025.11 | 53.9 | 53.4 | 54.5 | |
| FactLLaMAApproach category=LLM-based approach2026.04 | 53.76 | 53.76 | 54 | |
| CofCEDApproach category=Traditional approach2026.04 | 51.07 | 52.99 | 50.99 | |
| MUSERApproach category=Traditional approach2026.04 | 47.48 | 47.47 | 47.5 | |
| SBERT-FCApproach category=Traditional approach2026.04 | 45.51 | 51.06 | 45.92 | |
| GenFE-MTApproach category=Traditional approach2026.04 | 45.08 | 45.64 | 45.27 | |
| GenFEApproach category=Traditional approach2026.04 | 44.43 | 44.29 | 44.74 | |
| GPT-3.5claimApproach category=LLM-based approach2026.04 | 44.43 | 47.72 | 48.62 | |
| ChatGPTx -> y=c → v; exp, No Closed-Source API Dependency=false2025.11 | 44.43 | 47.72 | 48.62 | |
| dEFENDApproach category=Traditional approach2026.04 | 44.07 | 44.93 | 43.26 | |
| GPT-3.5fullApproach category=LLM-based approach2026.04 | 39.31 | 39.48 | 45.07 | |
| ChatGPTx -> y=c, evi → v; exp, No Closed-Source API Dependency=false2025.11 | 39.31 | 39.48 | 45.07 | |
| LLaMA2claimApproach category=LLM-based approach2026.04 | 36.77 | 37.3 | 38.03 |