Want to create an interactive transcript for this episode?
Podcast: Arxiv Papers
Episode: Using Attention Sinks to Identify and Evaluate Dormant Heads in Pretrained LLMs
Description: The paper introduces "dormant attention heads" in multi-head attention, analyzing their impact on model performance and revealing their early emergence and dependency on input text characteristics.https://arxiv.org/abs//2504.03889YouTube: https://www.youtube.com/@ArxivPapersTikTok: https://www.tiktok.com/@arxiv_papersApple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers