[CaCL] 1/18: Reinforced Self-Training (ReST) for Language Modeling

[Oh, Byung-Doh]

Hi everyone,

Next week, we'll discuss the following paper; I might follow up with additional resources that may be helpful for fully appreciating this paper as well.

Reinforced Self-Training (ReST) for Language Modeling
https://arxiv.org/pdf/2308.08998.pdf
Reinforcement learning from human feedback (RLHF) can improve the quality of large language model’s (LLM) outputs by aligning them with human preferences. We propose a simple algorithm for aligning LLMs with human preferences inspired by growing batch reinforcement learning (RL), which we call Reinforced Self-Training (ReST). Given an initial LLM policy, ReST produces a dataset by generating samples from the policy, which are then used to improve the LLM policy using offline RL algorithms. ReST is more efficient than typical online RLHF methods because the training dataset is produced offline, which allows data reuse. While ReST is a general approach applicable to all generative learning settings, we focus on its application to machine translation. Our results show that ReST can substantially improve translation quality, as measured by automated metrics and human evaluation on machine translation benchmarks in a compute and sample-efficient manner.

Best,
Byung-Doh

=================
Byung-Doh Oh (he/him/his)
Ph.D. Candidate
Department of Linguistics
The Ohio State University

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osu.edu/pipermail/cacl/attachments/20240111/1a669df3/attachment.html>