The paper introduces generative infinite-vocabulary transformers (GIVT) that generate vector sequences with real-valued entries, proposing modifications to decoder-only transformers. GIVT shows competitive results in image generation, causal modeling, panoptic segmentation, and depth estimation.

https://arxiv.org/abs//2312.02116

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