Language Models Might Not Understand You: Evaluating Theory of Mind via Story Prompting
cs.CL, cs.AI
Submitted: 2025-06-23
Updated: 2026-09-01
Comments: 25 pages, 19 figures
Code: https://github.com/ngetachew/StorySim
License: http://creativecommons.org/licenses/by/4.0/
The gist: We introduce StorySim, a programmable framework for synthetically generating stories to evaluate the theory of mind (ToM) and world modeling (WM) capabilities of large language models (LLMs).
Terminology
Abstract
We introduce StorySim, a programmable framework for synthetically generating stories to evaluate the theory of mind (ToM) and world modeling (WM) capabilities of large language models (LLMs). Unlike prior benchmarks that may suffer from contamination in pretraining data, or rely on an LLM for generation, StorySim produces novel, compositional story prompts anchored by a highly controllable Storyboard, enabling precise manipulation of character perspectives and events. Using StorySim, we evaluate LLMs across three kinds of ToM tasks: false belief, goal-directed, and preference attribution tasks. We then use our framework to design first- and second-order ToM tasks alongside WM tasks that control for the ability to track and model mental states. Our experiments across a suite of LLMs show that most models achieve higher accuracy on WM tasks than on ToM tasks, and that some models tend to reason more accurately when the subject of reasoning is a person rather than an inanimate object. Additionally, models struggle with goal-directed reasoning the most, and performance does not always scale with model size. All code for generating data and evaluations is freely available.
Sources
- GPT-4 Technical Report
- On the Diversity of Synthetic Data and its Impact on Training Large Language Models
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- Surveying the Effects of Quality, Diversity, and Complexity in Synthetic Data From Large Language Models
- GPT-4o System Card
- OpenAI o1 System Card
- Evaluating Large Language Models in Theory of Mind Tasks
- Do Large Language Models Have a Planning Theory of Mind? Evidence from MindGames: a Multi-Step Persuasion Task
- PHYBench: Holistic Evaluation of Physical Perception and Reasoning in Large Language Models
- Explore Theory of Mind: Program-guided adversarial data generation for theory of mind reasoning
- Does Synthetic Data Generation of LLMs Help Clinical Text Mining?
- Large Language Models Fail on Trivial Alterations to Theory-of-Mind Tasks
- MiMo: Unlocking the Reasoning Potential of Language Model -- From Pretraining to Posttraining
- Logic-RL: Unleashing LLM Reasoning with Rule-Based Reinforcement Learning
- Qwen3 Technical Report
Related papers
- Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
- Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
- Subliminal Steering: Stronger Encoding of Hidden Signals
- MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
- The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
- Untangling the Mechanisms of Misleading Context in Medical Question Answering