Large language models (LLMs) can be used as strong encoders for visual tasks without language by employing a frozen transformer block from pre-trained LLMs. This approach improves performance across various computer vision tasks and is supported by the information filtering hypothesis.

https://arxiv.org/abs//2310.12973

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