ToM-SSI: Evaluating Theory of Mind in Situated Social Interactions

2025-09-05

AI Summary

Existing Theory of Mind (ToM) benchmarks predominantly rely on the Sally-Anne paradigm and are confined to unimodal, single-turn textual dialogues, failing to capture complex, multimodal, multi-agent social interactions. Method: We introduce ToM-SSI—the first embodied, multi-agent ToM evaluation benchmark designed for situated social scenarios—featuring hybrid cooperative-competitive settings and parallel multi-agent mental state inference tasks. It supports joint visual-language-spatial modeling and trajectory-language co-reasoning. Contribution/Results: Empirical evaluation reveals that state-of-the-art large language and multimodal models exhibit significant deficiencies in multi-agent collaborative reasoning and in scenarios involving cooperation breakdown, exposing critical limitations in current ToM modeling. ToM-SSI establishes a novel, systematic paradigm for assessing embodied social cognition and provides a scalable, extensible testbed for advancing ToM capabilities in interactive, real-world environments.

Technology Category

Cognitive Modeling & Cognitive Systems: Social Cognition And InteractionMultiagent Systems: Modeling other AgentsMachine Learning: Large Multimodal Models (LMMs)

Application Category

Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactionsSearch and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAGSocial Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Most existing Theory of Mind (ToM) benchmarks for foundation models rely on variations of the Sally-Anne test, offering only a very limited perspective on ToM and neglecting the complexity of human social interactions. To address this gap, we propose ToM-SSI: a new benchmark specifically designed to test ToM capabilities in environments rich with social interactions and spatial dynamics. While current ToM benchmarks are limited to text-only or dyadic interactions, ToM-SSI is multimodal and includes group interactions of up to four agents that communicate and move in situated environments. This unique design allows us to study, for the first time, mixed cooperative-obstructive settings and reasoning about multiple agents' mental state in parallel, thus capturing a wider range of social cognition than existing benchmarks. Our evaluations reveal that the current models' performance is still severely limited, especially in these new tasks, highlighting critical gaps for future research.

Problem

• Evaluating Theory of Mind in complex social interactions
• Testing multimodal group interactions with spatial dynamics
• Assessing mental state reasoning for multiple parallel agents

Innovation

• Multimodal benchmark with group interactions
• Evaluates mixed cooperative-obstructive social settings
• Assesses parallel mental state reasoning for multiple agents