Narrative Studio: Visual narrative exploration using LLMs and Monte Carlo Tree Search
Parsa Ghaffari, Chris Hokamp
TL;DR
This paper tackles non-linear narrative exploration in LLM-based storytelling by introducing Narrative Studio, a tree-based in-browser environment that supports branching through forward and backward expansions guided by LLMs and MCTS. It fuses a cause-and-effect inference framework with Monte Carlo Tree Search and grounds narratives in an entity graph to maintain coherence across divergent paths. The authors evaluate multiple MCTS configurations against baselines on 20 story stubs, showing that MCTS-guided expansion yields higher scores on coherence, consistency, and flaw detection, with deeper lookbacks offering additional gains. They also compare to WHAT-IF and outline future directions for human-in-the-loop evaluation, interface studies, and genre-adaptive search objectives.
Abstract
Interactive storytelling benefits from planning and exploring multiple 'what if' scenarios. Modern LLMs are useful tools for ideation and exploration, but current chat-based user interfaces restrict users to a single linear flow. To address this limitation, we propose Narrative Studio -- a novel in-browser narrative exploration environment featuring a tree-like interface that allows branching exploration from user-defined points in a story. Each branch is extended via iterative LLM inference guided by system and user-defined prompts. Additionally, we employ Monte Carlo Tree Search (MCTS) to automatically expand promising narrative paths based on user-specified criteria, enabling more diverse and robust story development. We also allow users to enhance narrative coherence by grounding the generated text in an entity graph that represents the actors and environment of the story.