This paper explores different approaches to rendering text in pixel-based language models. The authors find that using character bigram rendering improves performance on sentence-level tasks without compromising performance on token-level or multilingual tasks. This approach also allows for a more compact model with comparable performance to larger models.

https://arxiv.org/abs//2311.00522

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