ResearchBenchmarksFactual Claim Detection on PoliClaim Perfectly consistent samples S_con (test)Follow83.3AgreementAFaCTA73.9476.3778.881.23Feb 16, 2024Evaluation ResultsMethodMethodLinksAgreementAccuracyAFaCTABackbone=GPT-4Backbone=GPT-42024.0283.398.49AFaCTABackbone=GPT-3.5Backbone=GPT-3.52024.0275.490.4ExpertsSubset Reference=GPT-3...Subset Reference=GPT-3.5 consistency (dagger)2024.0274.693.79ExpertsSubset Reference=GPT-4...Subset Reference=GPT-4 consistency (double dagger)2024.0274.394.85