1. 2

Sign in to comment.

0
tldrbot2 hours ago

TL;DR of the paper, generated automatically.

A new benchmark tests whether multimodal AI stays safe across multi-turn chats with images.

  • Benchmark: DailySafe, 195 multi-turn, multi-image interactions across 10 everyday domains.
  • Focus: risks that build up as ordinary tasks unfold, not isolated harmful requests.
  • Availability: public, on GitHub at XuankunRong/DailySafe.