aiqbench analysis
This research advances reinforcement learning by enabling agents to follow complex temporal logic instructions in sub-symbolic environments without assuming prior knowledge of symbol-to-observation mappings. By jointly training a policy and a neural reward machine, the approach achieves zero-shot generalization comparable to methods using true grounding while outperforming existing techniques that avoid such assumptions. The work is significant for improving multi-task learning capabilities where symbolic interpretations must be inferred directly from raw sensory data like vision inputs.
Grounding LTL Tasks in Sub-Symbolic RL Environments for Zero-Shot Generalization
arXiv:2602.09761v2 Announce Type: replace Abstract: In this work we address the problem of training a Reinforcement Learning agent to follow multiple temporally-extended instructions expressed in Linear Temporal Logic in sub-symbolic environments. Previous multi-task work has mostly relied on knowledge of the mapping between raw observations and symbols appearing in the formulae. We drop this unrealistic assumption by jointly training a multi-task policy and a symbol grounder with the same experience. The symbol grounder is trained only from raw observations and sparse rewards via Neural Reward Machines in a semi-supervised fashion. Experiments on vision-based environments show that our method achieves performance comparable to using the true symbol grounding and significantly outperforms the only other previous method for multi-task learning that does not assume knowledge of the true symbol grounding.
Abstract:In this work we address the problem of training a Reinforcement Learning agent to follow multiple temporally-extended instructions expressed in Linear Temporal Logic in sub-symbolic environments. Previous multi-task work has mostly relied on knowledge of the mapping between raw observations and symbols appearing in the formulae. We drop this unrealistic assumption by jointly training a multi-task policy and a symbol grounder with the same experience. The symbol grounder is trained only from raw observations and sparse rewards via Neural Reward Machines in a semi-supervised fashion. Experiments on vision-based environments show that our method achieves performance comparable to using the true symbol grounding and significantly outperforms the only other previous method for multi-task learning that does not assume knowledge of the true symbol grounding.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2602.09761 [cs.LG]
(or arXiv:2602.09761v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2602.09761 Focus to learn more arXiv-issued DOI via DataCite
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 · 218 words
Related coverage
Bounding Boxes as Goals: Language-Conditioned Grasping via Neuro-Symbolic Planning
arXiv · Jun 16, 2026
Bounding Boxes as Goals: Language-Conditioned Grasping via Neuro-Symbolic Planning
arXiv · Jun 12, 2026
Neuro-Symbolic Agentic AI for Networked Low-Altitude UAVs
arXiv · Sep 18, 2026
MotorMind: Scaffolding General Vision Language Models for Zero-Shot Robot Manipulation
arXiv · Sep 30, 2026