analiza aiqbench
Mimir introduces a neuro-symbolic architecture that explicitly separates world state from task progression, enabling embodied agents to dynamically ground goals in real-time environmental facts. This approach significantly boosts success rates on long-horizon tasks like EB-ALFRED and EB-Habitat by resolving the limitations of flat historical data or implicit policy states. The system demonstrates substantial performance gains over prior models while maintaining an open-source commitment through upcoming code releases.
Mimir: A Neuro-Symbolic Memory System with Dynamic Grounding for Embodied Agents in Interactive Environments
arXiv:2608.04933v1 Announce Type: new Abstract: Long-horizon embodied task requires agents to act under partial observability while preserving both scene belief and execution progress. Flat histories or implicit policy states may contain past observations, but they do not provide an explicit interface for deciding which world facts support the currently active goal. We introduce Mimir, a neuro-symbolic memory that separates world memory from task memory and dynamically grounds them before each action. World memory maintains object locations, object states, and perceptual evidence, while task memory maintains an ordered goal agenda, progress state, hand state, failures, and execution constraints. A grounding module binds the active goal to recalled world candidates, fills missing source locations, and attaches evidence before planning and embodiment-specific execution. Across tested backbones, Mimir consistently improves on different EB-ALFRED and EB-Habitat tasks, with maximum gains of 42.5% and average gains of 23.0%, respectively. Compared with the best results among prior agent and memory systems evaluated under the same backbone, Mimir improves the overall average success rate by 8.5%. Finally, on the EB-Habitat Long-horizon subset, Mimir achieves 86.0% success rate, substantially outperforming current closed-source models. Our code will be released soon.
Abstract:Long-horizon embodied task requires agents to act under partial observability while preserving both scene belief and execution progress. Flat histories or implicit policy states may contain past observations, but they do not provide an explicit interface for deciding which world facts support the currently active goal. We introduce Mimir, a neuro-symbolic memory that separates world memory from task memory and dynamically grounds them before each action. World memory maintains object locations, object states, and perceptual evidence, while task memory maintains an ordered goal agenda, progress state, hand state, failures, and execution constraints. A grounding module binds the active goal to recalled world candidates, fills missing source locations, and attaches evidence before planning and embodiment-specific execution. Across tested backbones, Mimir consistently improves on different EB-ALFRED and EB-Habitat tasks, with maximum gains of 42.5% and average gains of 23.0%, respectively. Compared with the best results among prior agent and memory systems evaluated under the same backbone, Mimir improves the overall average success rate by 8.5%. Finally, on the EB-Habitat Long-horizon subset, Mimir achieves 86.0% success rate, substantially outperforming current closed-source models. Our code will be released soon.
Robotics (cs.RO)
arXiv:2608.04933 [cs.RO]
(or arXiv:2608.04933v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2608.04933 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.
Text integral sub licența deschisă a publicației · 278 cuvinte
Subiecte conexe
Neuro-Symbolic Agentic AI for Networked Low-Altitude UAVs
arXiv · 18 sept. 2026
Academia is for Ambition — Alex Zhang, MIT
Latent Space · 2 oct. 2026
Neuro-Symbolic Meta-Policies for Temporal Knowledge-Graph Memory under Partial Observability
arXiv · 23 iul. 2026
Gamma: Grounded Multi-Agent Memory with Mechanical Audit for Memory-Intensive Manipulation
arXiv · 12 oct. 2026