SafeLens: Deliberate and Efficient Video Guardrails with Fast-and-Slow Screening

In three linesSafeLens introduces a two-tier video moderation architecture (fast-and-slow) to reduce inference costs. The framework filters SafeWatch dataset to 2.4% via influence-guided filtering and augments with Chain-of-Thought traces. It outperforms SafeWatch-8B, OmniGuard-7B, GPT-5.4, and Gemini-3.1-pro on real and AI-generated video benchmarks.

Summary generated by Claude — human-verified