Reproduce “Rethinking Content Moderation via Severity-Aware Ranking” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yang Ou, Jielei Zhang, Yu Dai, Zian Yang, Run Yang, Xiang Li, Weihang Wang, Xueqian Zhao and 2 more
KUROKO turns video moderation into severity ranking with rule-based reasons, rather than a yes-or-no label. It combines matched policy rules into a graded score and routes suspected content to a multimodal model for review. In a matched four-class test, its Macro-F1 was 0.5557 versus 0.4361 for an ordinal head; a live test improved outcomes at equal review volume, without directly measuring over-blocking.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.