mlsanityData to 7 Oct 2026

Neuro-symbolic Action Masking for Deep Reinforcement Learning

arXiv 2602.10598 ↗First posted 11 Feb 2026cs.AI cs.LGcited by 1

Not on the map. Why?

Abstract (arXiv)

Deep reinforcement learning (DRL) may explore infeasible actions during training and execution. Existing approaches assume a symbol grounding function that maps high-dimensional states to consistent symbolic representations and a manually specified action masking techniques to constrain actions. In this paper, we propose Neuro-symbolic Action Masking (NSAM), a novel framework that automatically learn symbolic models, which are consistent with given domain constraints of high-dimensional states, in a minimally supervised manner during the DRL process. Based on the learned symbolic model of states, NSAM learns action masks that rules out infeasible actions. NSAM enables end-to-end integration of symbolic reasoning and deep policy optimization, where improvements in symbolic grounding and policy learning mutually reinforce each other. We evaluate NSAM on multiple domains with constraints, and experimental results demonstrate that NSAM significantly improves sample efficiency of DRL agent while substantially reducing constraint violations.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 32 of its 72 references that are in the map. First posted on arXiv 11 Feb 2026.

Builds on

32 of 72 references.

References · 32

  1. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  2. 2017Proximal Policy Optimization Algorithmscited by 14,441
  3. 2019Dota 2 with Large Scale Deep Reinforcement Learningcited by 1,084
  4. 2017Rainbow: Combining Improvements in Deep Reinforcement Learningcited by 774
  5. 2017Safe Reinforcement Learning via Shieldingcited by 427
  6. 2018Reward Constrained Policy Optimizationcited by 397
  7. 2019Neural Probabilistic Logic Programming in DeepProbLogcited by 389
  8. 2015Risk-Constrained Reinforcement Learning with Percentile Risk Criteriacited by 320
  9. 2017StarCraft II: A New Challenge for Reinforcement Learningcited by 268
  10. 2017A Semantic Loss Function for Deep Learning with Symbolic Knowledgecited by 267
  11. 2020Reward Machines: Exploiting Reward Function Structure in Reinforcement Learningcited by 183
  12. 2019SATNet: Bridging deep learning and logical reasoning using a differentiable satisfiability solvercited by 173
20 more
  1. 2020A Closer Look at Invalid Action Masking in Policy Gradient Algorithmscited by 160
  2. 2019Mastering Complex Control in MOBA Games with Deep Reinforcement Learningcited by 150
  3. 2024SERL: A Software Suite for Sample-Efficient Robotic Reinforcement Learningcited by 113
  4. 2020Action Space Shaping in Deep Reinforcement Learningcited by 45
  5. 2020Neural-Symbolic Integration: A Compositional Perspectivecited by 45
  6. 2023Safe Reinforcement Learning via Probabilistic Logic Shieldscited by 41
  7. 2021Modular Design Patterns for Hybrid Learning and Reasoning Systems: a taxonomy, patterns and use casescited by 39
  8. 2021Feasible Actor-Critic: Constrained Reinforcement Learning for Ensuring Statewise Safetycited by 30
  9. 2023Safe Exploration in Reinforcement Learning: A Generalized Formulation and Algorithmscited by 27
  10. 2016SDDs are Exponentially More Succinct than OBDDscited by 19
  11. 2021Gym-μRTS: Toward Affordable Full Game Real-time Strategy Games Research with Deep Reinforcement Learningcited by 19
  12. 2024Excluding the Irrelevant: Focusing Reinforcement Learning through Continuous Action Maskingcited by 17
  13. 2023AutoCost: Evolving Intrinsic Cost for Zero-violation Reinforcement Learningcited by 16
  14. 2025Probabilistic Shielding for Safe Reinforcement Learningcited by 12
  15. 2025Handling Delay in Real-Time Reinforcement Learningcited by 7
  16. 2023Reinforcement Learning by Guided Safe Explorationcited by 5
  17. 2024Knowledge Transfer from Simple to Complex: A Safe and Efficient Reinforcement Learning Framework for Autonomous Driving Decision-Makingcited by 5
  18. 2025Learning from Less: Guiding Deep Reinforcement Learning with Differentiable Symbolic Planningcited by 3
  19. 2024Towards Real-World Efficiency: Domain Randomization in Reinforcement Learning for Pre-Capture of Free-Floating Moving Targets by Autonomous Robotscited by 2
  20. 2025CRLLK: Constrained Reinforcement Learning for Lane Keeping in Autonomous Drivingcited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Short-Term-to-Long-Term Memory Transfer for Knowledge Graphs under Partial Observabilitycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Few-Shot Neuro-Symbolic Imitation Learning for Long-Horizon Planning and Acting
  2. 2024BlendRL: A Framework for Merging Symbolic and Neural Policy Learning
  3. 2026NS-VLA: Towards Neuro-Symbolic Vision-Language-Action Models
  4. 2022Masked Autoencoding for Scalable and Generalizable Decision Making
  5. 2026Build on Priors: Vision--Language--Guided Neuro-Symbolic Imitation Learning for Data-Efficient Real-World Robot Manipulation
  6. 2025Compositional Concept-Based Neuron-Level Interpretability for Deep Reinforcement Learning
  7. 2025Hierarchical Neuro-Symbolic Decision Transformer
  8. 2025Masked IRL: LLM-Guided Reward Disambiguation from Demonstrations and Language