Can pre-trained language models generate titles for research papers?

TL;DR

This paper explores the use of pre-trained language models, including fine-tuned models and GPT-3.5-turbo, for automated research paper title generation from abstracts, showing promising results with PEGASUS-large.

Contribution

It introduces a method for fine-tuning pre-trained models to generate paper titles and compares their performance with GPT-3.5-turbo in zero-shot settings.

Findings

PEGASUS-large outperforms other models on most metrics

GPT-3.5-turbo can generate creative titles in zero-shot mode

AI-generated titles are generally accurate and appropriate

Abstract

The title of a research paper communicates in a succinct style the main theme and, sometimes, the findings of the paper. Coming up with the right title is often an arduous task, and therefore, it would be beneficial to authors if title generation can be automated. In this paper, we fine-tune pre-trained language models to generate titles of papers from their abstracts. Additionally, we use GPT-3.5-turbo in a zero-shot setting to generate paper titles. The performance of the models is measured with ROUGE, METEOR, MoverScore, BERTScore and SciBERTScore metrics. We find that fine-tuned PEGASUS-large outperforms the other models, including fine-tuned LLaMA-3-8B and GPT-3.5-turbo, across most metrics. We also demonstrate that ChatGPT can generate creative titles for papers. Our observations suggest that AI-generated paper titles are generally accurate and appropriate.