Ex3: Automatic Novel Writing by Extracting, Excelsior and Expanding

TL;DR

This paper introduces Ex3, a novel method for AI-generated long-form novels that extracts structural information from data, fine-tunes language models, and employs tree-like expansion to improve coherence and narrative quality.

Contribution

Ex3 combines structure extraction, dataset creation, and a tree-like expansion to enhance the quality and length of AI-generated novels, addressing coherence issues.

Findings

Ex3 produces higher-quality long novels than previous methods.

The method improves logical coherence and plot development.

It enables generation of arbitrarily long narratives.

Abstract

Generating long-term texts such as novels using artificial intelligence has always been a challenge. A common approach is to use large language models (LLMs) to construct a hierarchical framework that first plans and then writes. Despite the fact that the generated novels reach a sufficient length, they exhibit poor logical coherence and appeal in their plots and deficiencies in character and event depiction, ultimately compromising the overall narrative quality. In this paper, we propose a method named Extracting Excelsior and Expanding. Ex3 initially extracts structure information from raw novel data. By combining this structure information with the novel data, an instruction-following dataset is meticulously crafted. This dataset is then utilized to fine-tune the LLM, aiming for excelsior generation performance. In the final stage, a tree-like expansion method is deployed to facilitate the generation of arbitrarily long novels. Evaluation against previous methods showcases Ex3's ability to produce higher-quality long-form novels.