Evaluating Generative Models as Interactive Emergent Representations of Human-Like Collaborative Behavior

2026-05-05

AI Summary

This study investigates whether embodied foundation models spontaneously exhibit human-like collaborative behaviors and underlying theory-of-mind capabilities in cooperative settings. To this end, the authors design a 2D color-matching collaborative game environment where large language models (LLMs) interact with human users to complete coordination-dependent tasks. They introduce an evaluation framework grounded in five categories of collaborative behavior—such as perspective-taking and plan awareness—and employ an LLM-based adjudicator to automatically detect these behaviors, achieving high agreement with human annotations. Experimental results demonstrate that LLMs, despite lacking explicit collaborative training, consistently display human-like cooperative competencies. User studies further indicate that participants positively rated the LLMs’ task focus and proactiveness, though response latency and interaction naturalness remain areas for improvement.

Abstract

Human-AI collaboration requires AI agents to understand human behavior for effective coordination. While advances in foundation models show promising capabilities in understanding and showing human-like behavior, their application in embodied collaborative settings needs further investigation. This work examines whether embodied foundation model agents exhibit emergent collaborative behaviors indicating underlying mental models of their collaborators, which is an important aspect of effective coordination. This paper develops a 2D collaborative game environment where large language model agents and humans complete color-matching tasks requiring coordination. We define five collaborative behaviors as indicators of emergent mental model representation: perspective-taking, collaborator-aware planning, introspection, theory of mind, and clarification. An automated behavior detection system using LLM-based judges identifies these behaviors, achieving fair to substantial agreement with human annotations. Results from the automated behavior detection system show that foundation models consistently exhibit emergent collaborative behaviors without being explicitly trained to do so. These behaviors occur at varying frequencies during collaboration stages, with distinct patterns across different LLMs. A user study was also conducted to evaluate human satisfaction and perceived collaboration effectiveness, with the results indicating positive collaboration experiences. Participants appreciated the agents' task focus, plan verbalization, and initiative, while suggesting improvements in response times and human-like interactions. This work provides an experimental framework for human-AI collaboration, empirical evidence of collaborative behaviors in embodied LLM agents, a validated behavioral analysis methodology, and an assessment of collaboration effectiveness.

Problem

• human-AI collaboration
• embodied agents
• emergent behavior
• mental models
• foundation models

Innovation

• embodied foundation models
• emergent collaboration
• mental model representation
• LLM-based behavior detection
• human-AI coordination