From Personas to Plot: Character-Grounded Multi-Agent Story Generation for Long-Form Narratives

Jul 16, 2026

Upload by Anda

Long-form story generation with large language models often fails to keep characters consistent and maintain coherent plot lines, especially when multiple characters interact. This work presents Magnet, a multi-agent goal-driven narrative engine that uses persona-grounded character agents operating on a shared world state and evolving story goals. It pairs with Atlas, a graph-based hallucination detection pipeline that compares scene-level world representations across the story. Evaluations show improved coherence and fewer hallucinations than single-model prompting and IBSEN.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages22

Reading Time55 min

Abstract

Introduction

Related Work

Multi-Agent Narrative Generation

Structured State Tracking

arXiv :2607 .009 18v 1 [ cs .CL] 1 Jul 2026 From Personas to Plot: Character-Grounded Multi-Agent Story Generation for Long-Form Narratives Aayush Aluru* , Chloe Ho*1 , Muhammad Hammouri2 Kerry Luo3 , Myra Malik, Ryan Lagasse† Arjun Bahuguna†4 , Vasu Sharma† Pocket FM [aayush.aluru09@gmail.com](aayush.aluru09@gmail.com) , [ch4941@princeton.edu](ch4941@princeton.edu) [hammouri@umich.edu](hammouri@umich.edu) , [kerryluo1@gmail.com](kerryluo1@gmail.com) [arjunbahuguna251@gmail.com](arjunbahuguna251@gmail.com) Abstract Although large language models (LLMs) have demonstrated impressive creative fiction generation, they struggle to maintain narrative consistency and coherent plot lines in long-form stories. In this work, we introduce a unified framework for long-form narrative generation and verification. Magnet, a multi-agent goaldriven narrative engine for storytelling, generates stories with persona-grounded character agents that propose actions based on a shared world state and evolving story goals, while Atlas is a graph-based pipeline that compares scene-level world representations across a generated story to detect hallucinations. By evaluating Magnet using an LLM editor, pairwise rubric scoring, and Atlas, we show that our framework produces coherent narratives compared to single-model prompting and IBSEN. At 100 pages, Magnet reduced annotations and hallucinations by 41 and 50%, respectively, compared to the single model baseline and by 34 and 45%, respectively, compared to IBSEN, with pairwise rubric evaluation showing similar results. These results suggest that long-form narratives can emerge from explicit world-state tracking and goal-driven multi-agent generation, providing a foundation for controllable and structurally coherent long-form narrative generation. 1 Introduction Large language models (LLMs) have significantly advanced open-ended text generation, enabling their use for creating character personas and simulating complex interactions Wang et al. (2024b); OpenAI (2024) . Although LLMs have demonstrated strong narrative generation capabilities, they suffer from character inconsistency and plot discontinuity, limiting their ability to create high-quality long-form narratives Lu et al. (2026); Shao et al. (2023); Yao et al. (2019) . These failures become pronounced in multi-character environments, where LLMs struggle to balance narrative goals and character actions with complex relationships and interactions Park et al. (2023); Gao et al. (2024); Li et al. (2026a) . Recent work, including StoryVerse Wang et al. (2024a), Agents’Room Huot et al. (2025), and IBSEN Han et al. (2024) have explored the use of multi-agent systems in narrative generation, but they continue to rely on textual memory, limiting their ability to generate coherent stories. Lewis et al. (2021); Shinn et al. (2023); Liu et al. (2026); Teleki et al. (2025) . * Equal contribution. 1Princeton University. 2University of Michigan. 3University of Maryland.†Senior author. 4Universitat Pompeu Fabra. In addition to creative content generation, there remains a need to thoroughly evaluate this content. Existing work has made progress on long-form narrative understanding and factuality (Kočiský et al., 2018 ; Kim et al., 2023 ; Sansford et al., 2024 ; Lyu et al., 2025 ; Hamilton et al., 2025 ; Li et al., 2026a; Wu et al., 2025 ; Que et al., 2024) . However, these methods do not provide a framework for identifying hallucinations within long-form generated narratives. In this work, we introduce Magnet, a multi-agent story generation system to develop coherent long-form narratives and Atlas, a graph-based hallucination evaluation pipeline that identifies inconsistencies by comparing the world state representation of the present scene against those of previous scenes (Tian et al., 2026) . Our work aims to address the research question: Can long-form narratives emerge from interactions between character personas and updating

FAQs

What is Atlas and how does it evaluate hallucinations in generated stories?

Atlas is a graph-based evaluation pipeline that compares scene-level world state representations across a generated story to detect inconsistencies, providing interpretable signals of model failures.

Jul 16, 2026

Upload by Anda