Reproduce “Can Foundation Models Moderate Online Content? Evaluating Instruction- vs. Example-Driven Policy Operationalization” 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
Ayan Majumdar, Shounak Paul, Pushpdeep Singh, Ines Abdelaziz, Sayeh Jarollahi, Seungeon Lee, Krishna P. Gummadi, Ingmar Weber and 1 more
The authors test policy-guided and example-guided vision-language models for Bluesky content moderation on a human-reviewed benchmark. They find that on Random Posts Gemini 3.5 achieves F1 0.60, versus 0.22 for Bluesky’s moderation service. In a semantically selected near-moderated subset, 34.6% were judged unsafe although the service flagged none; this is not a platform-wide miss rate.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.