Narrative Theory-Driven LLM Methods for Automatic Story Generation and Understanding: A Survey

2026-06-18 · Source: cs.CL updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Social Sciences & Behavioral Studies · Depth: Expert, extended

Summary

This survey examines the application of narrative theory-driven large language model (LLM) methods for automatic story generation and understanding. It proposes a taxonomy for ongoing efforts, reflecting established distinctions in narratology, and analyzes patterns across narrative datasets, tasks, theories, NLP pipelines, and methodological trends like prompting and fine-tuning. The authors highlight LLMs' ability to easily connect NLP pipelines with abstract narrative concepts, fostering interdisciplinary collaboration. A key challenge identified is the absence of a unified definition or benchmark for narrative-related tasks, which complicates model comparison. Future directions emphasize defining and improving theory-based metrics for individual narrative attributes, conducting large-scale theory-driven literary/social/cultural analysis, and creating experiments to validate or refine narrative theories. This work provides a foundational overview for more systematic and theoretically informed narrative research in NLP.

Key takeaway

For AI Scientists and NLP Engineers developing story generation or understanding systems, you should integrate specific narrative theories beyond classical narratology to address current limitations. Focus on developing precise, theory-based metrics for individual narrative attributes rather than pursuing a single, generalized "narrative quality" benchmark. Consider utilizing LLMs for large-scale annotation and validation of specific narrative theories, and explore post-training methods for story generation to potentially create "narration-tuned" LLMs.

Key insights

LLMs bridge abstract narrative theories with NLP, enabling scalable story generation and understanding despite definitional challenges.

Principles

• Narrative research benefits from interdisciplinary theory.
• LLMs enable applying abstract narrative concepts.
• Focus on specific metrics over general benchmarks.

Method

The survey proposes a taxonomy for LLM-based narrative research, categorizing efforts by narrative levels (fabula, discourse, narration) and theoretical focus (classical, cognitive, contextual narratology) within the NLP pipeline (labeling, modeling, evaluation).

In practice

• Use LLMs for large-scale, theory-driven text annotation.
• Employ multi-agent systems for complex story generation.
• Fine-tune smaller models for specific narrative tasks.

Topics

• Narrative Theory
• Large Language Models
• Story Generation
• Story Understanding
• Natural Language Processing
• Narratology
• Prompt Engineering

Best for: AI Scientist, NLP Engineer, Research Scientist

Related on AIssential

• StoryCoder: Narrative Reformulation for Structured Reasoning in LLM Code Generation — Narrative reformulation of code problems significantly improves LLM code generation by structuring reasoning.
• Do Large Language Models Always Tell The Same Stories? — Large Language Models consistently produce narratives that are more homogeneous than human-written stories, even with common mitigation efforts.
• LLMs for automatic annotation of Mandarin narrative transcripts — LLMs can automate complex discourse-level annotation in non-English speech, significantly reducing time.
• Retell, Reward, Repeat: Reinforcement Learning for Narrative Theory-Informed Story Retelling — Reinforcement learning with narrative theory-informed AI feedback improves story diversity and human narrative alignment.
• Narrative Knowledge Weaver: Narrative-Centric Retrieval-Augmented Reasoning for Long-Form Text Understanding — Narrative Knowledge Weaver (NKW) enhances RAG by modeling evolving story worlds through structured narrative assets and multi-channel evidence assembly.
• SciText2Eq: Assessing LLMs for Explainable Equation Generation for Scientific Creativity — LLMs moderately generate equations from text but struggle with semantic accuracy and reliable self-evaluation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CL updates on arXiv.org.