We recommend to use the more recent benchmark, VeriTaS (ACL 2026 Oral), which fixes many weaknesses of previous fact-checking benchmarks, incl. ClaimReview2024+.

ClaimReview2024+ Benchmark

This is the ClaimReview2024+ (CR+) benchmark, a dataset used to evaluate multimodal automated fact-checking systems. The task is to classify each claim as either supported, refuted, misleading, or not enough information.… See the full description on the dataset page: https://huggingface.co/datasets/MAI-Lab/ClaimReview2024plus.