List your company
ResearchFormerly WizWand

Fact-checking on HOVER

71.82Macro F1 (2-hop)FOLK
49.834455.542261.2566.9578Jan 23, 2026
Updated 8mo ago

Evaluation results

MethodLinks
FOLKMethod Category=LLM-based reasoning methods, Source=Reproduction, Backbone Model=DeepSeek-Coder, k=10, threshold=0.92026.0171.8261.354.89——
ProgramFCMethod Category=LLM-based reasoning methods, Source=Reproduction, Backbone Model=DeepSeek-Coder, k=10, threshold=0.92026.0169.7860.2756.41——
RRCMethod Category=LLM-based reasoning methods, Backbone Model=DeepSeek-Coder, k=10, threshold=0.92026.0169.3363.9860.95——
FLAN-T5Method Category=LLM-based ICL methods, Source=Reported in [9], k=10, threshold=0.92026.0169.0260.2355.42——
DeBERTaV3-NLIMethod Category=Finetuned methods, Source=Reported in [9], k=10, threshold=0.92026.0168.7260.7656——
BootstrappingMethod Category=LLM-based reasoning methods, Source=Reproduction, Backbone Model=DeepSeek-Coder, k=10, threshold=0.92026.0168.4561.2959.23——
BiDevMethod Category=LLM-based reasoning methods, Source=Reproduction, Backbone Model=DeepSeek-Coder, k=10, threshold=0.92026.0167.8562.0358.87——
CodexMethod Category=LLM-based ICL methods, Source=Reported in [9], k=10, threshold=0.92026.0165.0756.6357.27——
RoBERTa-NLIMethod Category=Finetuned methods, Source=Reported in [9], k=10, threshold=0.92026.0163.6253.9952.4——
MULTIVERSMethod Category=Finetuned methods, Source=Reported in [9], k=10, threshold=0.92026.0160.1752.5551.86——
LisT5Method Category=Pretrained methods, Source=Reported in [9], k=10, threshold=0.92026.0152.5651.8950.46——
BERT-FCMethod Category=Pretrained methods, Source=Reported in [9], k=10, threshold=0.92026.0150.6849.8648.57——
Factcheck-GPTLLM=GPT-4o2026.01———4258
Factcheck-GPTLLM=GPT-4o-mini2026.01———2252
FactoolLLM=GPT-4o2026.01———4456
FactoolLLM=GPT-4o-mini2026.01———4151
FIRELLM=GPT-4o2026.01———3966
FIRELLM=GPT-4o-mini2026.01———3070
PCCLLM=GPT-4o2026.01———5276
PCCLLM=GPT-4o-mini2026.01———4571
SAFELLM=GPT-4o2026.01———4468
SAFELLM=GPT-4o-mini2026.01———2868