Benchmark shards for LureBench — detecting AI-generated fraud lures
(phishing, BEC, romance / pig-butchering). Version v1.2.
Each record follows the LureBench schema (see the
code repository): text, label
(fraud vs. benign), source (ai vs. human), typology, generator,
language, channel, persuasion.
Records: 20388 · fraud ratio: 0.3908 · AI ratio: 0.0289
Typologies: benign=12420, phishing=7527, bec=144, pig_butchering=148… See the full description on the dataset page:
https://huggingface.co/datasets/immu4989/lurebench-core.
