Saved in:
Bibliographic Details
Main Authors: Pimenta, Rui A., Schlippe, Tim, Schaaff, Kristina
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2508.16705
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916912295837696
author Pimenta, Rui A.
Schlippe, Tim
Schaaff, Kristina
author_facet Pimenta, Rui A.
Schlippe, Tim
Schaaff, Kristina
contents We investigate consciousness-like behaviors in Large Language Models (LLMs) using the Maze Test, challenging models to navigate mazes from a first-person perspective. This test simultaneously probes spatial awareness, perspective-taking, goal-directed behavior, and temporal sequencing-key consciousness-associated characteristics. After synthesizing consciousness theories into 13 essential characteristics, we evaluated 12 leading LLMs across zero-shot, one-shot, and few-shot learning scenarios. Results showed reasoning-capable LLMs consistently outperforming standard versions, with Gemini 2.0 Pro achieving 52.9% Complete Path Accuracy and DeepSeek-R1 reaching 80.5% Partial Path Accuracy. The gap between these metrics indicates LLMs struggle to maintain coherent self-models throughout solutions -- a fundamental consciousness aspect. While LLMs show progress in consciousness-related behaviors through reasoning mechanisms, they lack the integrated, persistent self-awareness characteristic of consciousness.
format Preprint
id arxiv_https___arxiv_org_abs_2508_16705
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Assessing Consciousness-Related Behaviors in Large Language Models Using the Maze Test
Pimenta, Rui A.
Schlippe, Tim
Schaaff, Kristina
Computation and Language
Artificial Intelligence
We investigate consciousness-like behaviors in Large Language Models (LLMs) using the Maze Test, challenging models to navigate mazes from a first-person perspective. This test simultaneously probes spatial awareness, perspective-taking, goal-directed behavior, and temporal sequencing-key consciousness-associated characteristics. After synthesizing consciousness theories into 13 essential characteristics, we evaluated 12 leading LLMs across zero-shot, one-shot, and few-shot learning scenarios. Results showed reasoning-capable LLMs consistently outperforming standard versions, with Gemini 2.0 Pro achieving 52.9% Complete Path Accuracy and DeepSeek-R1 reaching 80.5% Partial Path Accuracy. The gap between these metrics indicates LLMs struggle to maintain coherent self-models throughout solutions -- a fundamental consciousness aspect. While LLMs show progress in consciousness-related behaviors through reasoning mechanisms, they lack the integrated, persistent self-awareness characteristic of consciousness.
title Assessing Consciousness-Related Behaviors in Large Language Models Using the Maze Test
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2508.16705