LLMs process text with tokens, but individual tokens may not relate to word meanings. This study explores how LLMs convert tokens into higher-level representations, revealing an "erasure" effect.

https://arxiv.org/abs//2406.20086

YouTube: https://www.youtube.com/@ArxivPapers

TikTok: https://www.tiktok.com/@arxiv_papers

Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016

Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers