A Notion of Complexity for Theory of Mind via Discrete World Models

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Huang, X. Angelo, La Malfa, Emanuele, Marro, Samuele, Asperti, Andrea, Cohn, Anthony, Wooldridge, Michael
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866909341491134464
author Huang, X. Angelo
La Malfa, Emanuele
Marro, Samuele
Asperti, Andrea
Cohn, Anthony
Wooldridge, Michael
author_facet Huang, X. Angelo
La Malfa, Emanuele
Marro, Samuele
Asperti, Andrea
Cohn, Anthony
Wooldridge, Michael
contents Theory of Mind (ToM) can be used to assess the capabilities of Large Language Models (LLMs) in complex scenarios where social reasoning is required. While the research community has proposed many ToM benchmarks, their hardness varies greatly, and their complexity is not well defined. This work proposes a framework inspired by cognitive load theory to measure the complexity of ToM tasks. We quantify a problem's complexity as the number of states necessary to solve it correctly. Our complexity measure also accounts for spurious states of a ToM problem designed to make it apparently harder. We use our method to assess the complexity of five widely adopted ToM benchmarks. On top of this framework, we design a prompting technique that augments the information available to a model with a description of how the environment changes with the agents' interactions. We name this technique Discrete World Models (DWM) and show how it elicits superior performance on ToM tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2406_11911
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle A Notion of Complexity for Theory of Mind via Discrete World Models
Huang, X. Angelo
La Malfa, Emanuele
Marro, Samuele
Asperti, Andrea
Cohn, Anthony
Wooldridge, Michael
Artificial Intelligence
Computation and Language
Machine Learning
Theory of Mind (ToM) can be used to assess the capabilities of Large Language Models (LLMs) in complex scenarios where social reasoning is required. While the research community has proposed many ToM benchmarks, their hardness varies greatly, and their complexity is not well defined. This work proposes a framework inspired by cognitive load theory to measure the complexity of ToM tasks. We quantify a problem's complexity as the number of states necessary to solve it correctly. Our complexity measure also accounts for spurious states of a ToM problem designed to make it apparently harder. We use our method to assess the complexity of five widely adopted ToM benchmarks. On top of this framework, we design a prompting technique that augments the information available to a model with a description of how the environment changes with the agents' interactions. We name this technique Discrete World Models (DWM) and show how it elicits superior performance on ToM tasks.
title A Notion of Complexity for Theory of Mind via Discrete World Models
topic Artificial Intelligence
Computation and Language
Machine Learning
url https://arxiv.org/abs/2406.11911