Reproduce “CaseLaw: A Policy-Decoupled Framework for Training-Free Content Moderation” 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
Qi Lu, Sen Li, Likun Shi, Xiaolong Jiang
CaseLaw adapts content moderation to policy changes by updating an external library of precedent cases rather than changing deployed models. NoteCap preserves risk-bearing evidence in a structured representation, and retrieved cases vote for current labels when similarity meets case-specific thresholds. On MFO-RedNotes, post-change Safe-class precision and recall are 86.8% and 88.0%, versus 68.2% and 69.3% for Class-RAG, within this constructed setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.