Large language models require abundant compute and data for training, which is infeasible due to costs and data scarcity. The proposed WRAP method uses rephrased web data to improve pre-training efficiency and model performance.

https://arxiv.org/abs//2401.16380

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