Large language models (LLMs) exhibit a failure of generalization known as the Reversal Curse, where they struggle to answer questions in the reverse direction of their training data. This failure persists across different model sizes and families and is not improved by data augmentation.

https://arxiv.org/abs//2309.12288

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

PODCASTS:

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

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