Back to the Question: Tracing Visual Distraction Through Text Tokenswith Regev Cohen, Serena Yeung-Levy, Sivan Doveh1 versionW_3ns3cmjmv1 · current