Collective Critics for Creative Story Generation
Minwook Bae, Hyounghun Kim
2024-10-10
Summary
This paper discusses a new framework called CritiCS, which enhances the creative generation of long stories using Large Language Models (LLMs) by incorporating a collective critique process.
What's the problem?
Creating long, coherent stories with LLMs has been difficult because existing methods often focus too much on keeping the story consistent and logical, while neglecting creativity and expressiveness. This can lead to stories that are technically sound but lack the engaging qualities that captivate readers.
What's the solution?
The authors propose the CritiCS framework, which includes two main stages: CrPlan for refining story plans and CrText for generating the actual story. In each stage, a group of LLM critics works together with a leader to review and improve drafts through multiple rounds of feedback. This collaborative approach helps enhance creativity and reader engagement while still maintaining a coherent narrative.
Why it matters?
This research is important because it shows how collaboration between LLMs can lead to more creative and engaging storytelling. By allowing human writers to participate in the critique process, CritiCS promotes interactive human-machine collaboration in writing, potentially transforming how stories are generated and experienced.
Abstract
Generating a long story of several thousand words with narrative coherence using Large Language Models (LLMs) has been a challenging task. Previous research has addressed this challenge by proposing different frameworks that create a story plan and generate a long story based on that plan. However, these frameworks have been mainly focusing on maintaining narrative coherence in stories, often overlooking creativity in story planning and the expressiveness of the stories generated from those plans, which are desirable properties to captivate readers' interest. In this paper, we propose Collective Critics for Creative Story Generation framework (CritiCS), which is composed of plan refining stage (CrPlan) and story generation stage (CrText), to integrate a collective revision mechanism that promotes those properties into long-form story generation process. Specifically, in each stage, a group of LLM critics and one leader collaborate to incrementally refine drafts of plan and story throughout multiple rounds. Extensive human evaluation shows that the CritiCS can significantly enhance story creativity and reader engagement, while also maintaining narrative coherence. Furthermore, the design of the framework allows active participation from human writers in any role within the critique process, enabling interactive human-machine collaboration in story writing.