aiqbench analysis
This research introduces a neurosymbolic agent architecture that combines vision-language models with constrained symbolic planning to ensure the executability of long-horizon household tasks. By integrating Monte Carlo tree search with PDDL transition models, the system achieves over 95% success rates in simulated environments while significantly reducing token usage compared to standard generative approaches. The study demonstrates a critical shift from generating plausible but potentially invalid plans to constructing actions that are guaranteed correct under defined environment dynamics.
Neurosymbolic Embodied Agents
arXiv:2608.16794v1 Announce Type: new Abstract: Language and vision-language models generate plausible embodied plans but do not guarantee executability, as their outputs can violate environment dynamics or act on incorrectly grounded entities. We present a neurosymbolic agent that factors long-horizon household tasks into task-directed visual exploration and constrained symbolic planning. In the first phase, a vision-language model and exploration harness acquire goal-relevant predicates and instance bindings from egocentric observations and grounded interactions, producing a symbolic initial state. In the second, a PDDL transition model restricts decoding to tokens that extend applicable actions. Monte Carlo tree search then evaluates executable continuations using a domain-independent planning heuristic. The resulting plans are executable by construction under the transition model, with transfer to the environment conditioned on correct visual grounding. On VirtualHome and ALFWorld, open 4B-27B models exceed 90% success in both environments, and our smallest agent substantially outperforms a 27B direct visual policy in each. Constraints and search prove complementary rather than interchangeable: in ALFWorld either alone solves under a third of tasks, whereas their combination solves over 95%. The method also uses several times fewer generated tokens than extended thinking and far fewer model-visible images than direct interaction, and residual failures localize to state acquisition rather than plan generation without any specialized training.
Abstract:Language and vision-language models generate plausible embodied plans but do not guarantee executability, as their outputs can violate environment dynamics or act on incorrectly grounded entities. We present a neurosymbolic agent that factors long-horizon household tasks into task-directed visual exploration and constrained symbolic planning. In the first phase, a vision-language model and exploration harness acquire goal-relevant predicates and instance bindings from egocentric observations and grounded interactions, producing a symbolic initial state. In the second, a PDDL transition model restricts decoding to tokens that extend applicable actions. Monte Carlo tree search then evaluates executable continuations using a domain-independent planning heuristic. The resulting plans are executable by construction under the transition model, with transfer to the environment conditioned on correct visual grounding. On VirtualHome and ALFWorld, open 4B-27B models exceed 90% success in both environments, and our smallest agent substantially outperforms a 27B direct visual policy in each. Constraints and search prove complementary rather than interchangeable: in ALFWorld either alone solves under a third of tasks, whereas their combination solves over 95%. The method also uses several times fewer generated tokens than extended thinking and far fewer model-visible images than direct interaction, and residual failures localize to state acquisition rather than plan generation without any specialized training.
Robotics (cs.RO); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
arXiv:2608.16794 [cs.RO]
(or arXiv:2608.16794v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2608.16794 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 305 words
Related coverage
If, Then, Otherwise: Diagnosing Conditional Branching in Vision-Language Navigation
arXiv · Aug 19, 2026
Humanity’s Last Invention — Richard Socher of Recursive
Latent Space · Sep 14, 2026
Learning to Assemble Novel Structures with Unfamiliar Parts under Semantic Constraints
arXiv · Aug 17, 2026
How to Train a Cross-Embodiment Robot Navigation Policy with AI Agents
NVIDIA · Aug 26, 2026