Selecting Better Samples from Pre-trained LLMs: A Case Study on Question Generation
Abstract
Large Language Models (LLMs) have in recent years demonstrated impressive prowess in natural language generation. A common practice to improve generation diversity is to sample multiple outputs from the model. However, there lacks a simple and robust way of selecting the best output from these stochastic samples. As a case study framed in the context of question generation, we propose two prompt-based approaches to selecting high-quality questions from a set of LLM-generated candidates. Our method works under the constraints of 1) a black-box (non-modifiable) question generation model and 2) lack of access to human-annotated references -- both of which are realistic limitations for real-world deployment of LLMs. With automatic as well as human evaluations, we empirically demonstrate that our approach can effectively select questions of higher qualities than greedy generation.
Cite
@article{arxiv.2209.11000,
title = {Selecting Better Samples from Pre-trained LLMs: A Case Study on Question Generation},
author = {Xingdi Yuan and Tong Wang and Yen-Hsiang Wang and Emery Fine and Rania Abdelghani and Pauline Lucas and Hélène Sauzéon and Pierre-Yves Oudeyer},
journal= {arXiv preprint arXiv:2209.11000},
year = {2022}
}