Boosting large-language models for fact-checking: leveraging verbalized tabular data as evidence

Open paper
Benchmark Model Rank Results
table-based-fact-verification-on-tabfactTabV4FC (Qwen 2.5 - 72B)#18Test: 57.35
table-claim-verification-on-scitabTabV4FC (Qwen 2.5 - 72B)#1Macro-F1 (3-class): 0.4520
table-claim-verification-on-scitabTabV4FC (Qwen 2.5 - 14B)#4Macro-F1 (3-class): 0.3743
table-claim-verification-on-scitabTabV4FC (LLaMA 3.3 - 70B)#5Macro-F1 (3-class): 0.3449
table-claim-verification-on-scitabTabV4FC (DeepSeekR1-Qwen 32B)#6Macro-F1 (3-class): 0.3223
table-claim-verification-on-scitabTabV4FC (DeepSeekR1-LLaMA 70B)#7Macro-F1 (3-class): 0.2698
table-claim-verification-on-scitabTabV4FC (Qwen 2.5 - 32B)#8Macro-F1 (3-class): 0.1739

Rank counts only results with a code link.