Large Language Models Generate Harmful Content Using a Distinct Unified Mechanism (Hadas Orgad)

Round-17 hunt: picked; source: new-source search (r17: learned societies, museum/library lecture series, summer schools, scholarly conferences, newly covered languages); channel screens (sellers/exam/re-upload/devotional/vendor/events), provenance check, vendor/seller/devotional recheck and quality tier A passed; strict per-title check + final strict pass; oEmbed verified.