{"id":"9d0f7b54-5973-4b0a-87f0-b5a2923acf94","arxiv_id":"2411.13572","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"The PHAD dataset of 5,730 tobacco videos with rich metadata is new, but its classification results likely rely on text labels rather than visual understanding.","lead":"Researchers collected 5,730 tobacco-related videos from TikTok and YouTube, along with engagement stats, descriptions, and search keywords, and tested a vision-language classifier on them. The dataset is intended as a public health research resource, but the paper's performance claims appear inflated by using text that may give away the answer.","discovery_kind":"new_application","skeptic_critique":null,"referee_report":null,"author_rebuttal":null,"desk_editor":null,"rs_alignment":null,"lean_confirmation":null,"pith_extraction":null,"created_at":"2026-08-12T22:03:33.596329+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":null,"supporting_citations":[],"review_version":1}