Climate-FEVER is a dataset for verification of real-world climate change–related claims, built by adapting the FEVER evidence-retrieval methodology to claims collected from the web with input from climate scientists. BEIR (Benchmarking-IR) includes Climate-FEVER as one of its heterogeneous zero-shot information retrieval tasks: given a claim (query), a system must rank passages from a large Wikipedia-derived… See the full description on the dataset page: https://huggingface.co/datasets/orgrctera/beir_climate_fever.
