The paper proposes a method called Reinforced Fine-Tuning (ReFT) to enhance the generalizability of Large Language Models (LLMs) for reasoning tasks, using math problem-solving as an example. ReFT combines Supervised Fine-Tuning (SFT) with reinforcement learning and outperforms SFT in experiments.

https://arxiv.org/abs//2401.08967

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