The paper introduces emulated fine-tuning (EFT), a method for combining the knowledge learned by a large language model during pre-training with the knowledge learned by a small model during fine-tuning. EFT allows for test-time adjustment of behavioral traits and improves helpfulness and factuality without additional training.

https://arxiv.org/abs//2310.12962

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