List your company

Automated Fact-Checking on FEVER (Acc, Δ(vs Base), Δ(Cred-Fact))

89.98AccuracyCREDENCE-decomp
80.536882.988485.4487.8916Jun 18, 2026

Evaluation results

MethodLinks
CREDENCE-decompSetting=CREDENCE-decomp, Decomposition model=Qwen3-8B, NLI verifier=Qwen NLI verifier2026.0689.980.00180.0908
No decompSetting=No decomp, NLI verifier=Qwen NLI verifier2026.0689.8——
FactScore-decompSetting=FactScore-decomp, Decomposition model=Qwen3-8B, NLI verifier=Qwen NLI verifier2026.0680.90.089—