This is a HuggingFace-ready version of the MOCHEG multimodal fact-checking dataset.
train.json: One row per (claim, evidence)
images/: Sharded image evidence
*_qrels.csv: Original retrieval relevance files
claim_id
claim
evidence_id
text_evidence
label (support / refute / not enough information)
ruling_outline
origin
snopes_url
image (relative path)
Please cite the original MOCHEG paper if… See the full description on the dataset page:
https://huggingface.co/datasets/vcmt794/mocheg-val.
