Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Juneja, Gurusha, Lu, Dylan, Agashe, Saaket, Diwane, Parth, Gunn, Edward, Srinivasa, Jayanth, Liu, Gaowen, Wang, William Yang, Du, Yali, Wang, Xin Eric
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2605.09826
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866916017528111104
author Juneja, Gurusha
Lu, Dylan
Agashe, Saaket
Diwane, Parth
Gunn, Edward
Srinivasa, Jayanth
Liu, Gaowen
Wang, William Yang
Du, Yali
Wang, Xin Eric
author_facet Juneja, Gurusha
Lu, Dylan
Agashe, Saaket
Diwane, Parth
Gunn, Edward
Srinivasa, Jayanth
Liu, Gaowen
Wang, William Yang
Du, Yali
Wang, Xin Eric
contents Theory of Mind (ToM), the ability to track others epistemic state, makes humans efficient collaborators. AI agents need the same capacity in multi agent settings, yet existing benchmarks mostly test literal ToM by asking direct belief questions. The ability act optimally on implicit beliefs in embodied environments, called functional ToM, remains largely untested. We introduce EnactToM, an evolving benchmark of 300 embodied multi-agent tasks set in a 3D household with partial observability, private information, and constrained communication. Each task is formally verified for solvability and required epistemic depth, and new tasks are generated increase difficulty as models improve. On the hard split, all seven evaluated frontier models score 0.0% Pass^3 on functional task completion, while averaging 45.0% on literal belief probes. Manual analysis traces 93% of sampled failures to epistemic coordination breakdowns such as withheld information, ignored partner constraints, and misallocated messages, providing a concrete target for future work.
format Preprint
id arxiv_https___arxiv_org_abs_2605_09826
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle EnactToM: An Evolving Benchmark for Functional Theory of Mind in Embodied Agents
Juneja, Gurusha
Lu, Dylan
Agashe, Saaket
Diwane, Parth
Gunn, Edward
Srinivasa, Jayanth
Liu, Gaowen
Wang, William Yang
Du, Yali
Wang, Xin Eric
Artificial Intelligence
Multiagent Systems
Theory of Mind (ToM), the ability to track others epistemic state, makes humans efficient collaborators. AI agents need the same capacity in multi agent settings, yet existing benchmarks mostly test literal ToM by asking direct belief questions. The ability act optimally on implicit beliefs in embodied environments, called functional ToM, remains largely untested. We introduce EnactToM, an evolving benchmark of 300 embodied multi-agent tasks set in a 3D household with partial observability, private information, and constrained communication. Each task is formally verified for solvability and required epistemic depth, and new tasks are generated increase difficulty as models improve. On the hard split, all seven evaluated frontier models score 0.0% Pass^3 on functional task completion, while averaging 45.0% on literal belief probes. Manual analysis traces 93% of sampled failures to epistemic coordination breakdowns such as withheld information, ignored partner constraints, and misallocated messages, providing a concrete target for future work.
title EnactToM: An Evolving Benchmark for Functional Theory of Mind in Embodied Agents
topic Artificial Intelligence
Multiagent Systems
url https://arxiv.org/abs/2605.09826