mlsanityData to 7 Oct 2026

Learning with Logical Constraints but without Shortcut Satisfaction

arXiv 2403.00329 ↗First posted 1 Mar 2024cs.AI cs.LGcited by 10

Not on the map. Why?

Abstract (arXiv)

Recent studies in neuro-symbolic learning have explored the integration of logical knowledge into deep learning via encoding logical constraints as an additional loss function. However, existing approaches tend to vacuously satisfy logical constraints through shortcuts, failing to fully exploit the knowledge. In this paper, we present a new framework for learning with logical constraints. Specifically, we address the shortcut satisfaction issue by introducing dual variables for logical connectives, encoding how the constraint is satisfied. We further propose a variational framework where the encoded logical constraint is expressed as a distributional loss that is compatible with the model's original training loss. The theoretical analysis shows that the proposed approach bears salient properties, and the experimental evaluations demonstrate its superior performance in both model generalizability and constraint satisfaction.

Key facts. Cited by 10 papers in the map (data through 7 Oct 2026). Builds on 36 of its 68 references that are in the map. First posted on arXiv 1 Mar 2024.

Builds on

36 of 68 references.

References · 36

  1. 2021Convex optimizationcited by 2,479
  2. 2016Variational Inference: A Review for Statisticianscited by 1,466
  3. 2017Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semanticscited by 1,345
  4. 2018Multi-Task Learning as Multi-Objective Optimizationcited by 809
  5. 2019Neural Probabilistic Logic Programming in DeepProbLogcited by 389
  6. 2017A Semantic Loss Function for Deep Learning with Symbolic Knowledgecited by 267
  7. 2019On Gradient Descent Ascent for Nonconvex-Concave Minimax Problemscited by 266
  8. 2017Learning Explanatory Rules from Noisy Datacited by 210
  9. 2020Logic Tensor Networkscited by 197
  10. 2021DC3: A learning method for optimization with hard constraintscited by 181
  11. 2019SATNet: Bridging deep learning and logical reasoning using a differentiable satisfiability solvercited by 173
  12. 2017End-to-End Differentiable Provingcited by 159
24 more
  1. 2019A Meta-Transfer Objective for Learning to Disentangle Causal Mechanismscited by 158
  2. 2019Neural-Symbolic Computing: An Effective Methodology for Principled Integration of Machine Learning and Reasoningcited by 149
  3. 2019Neural Logic Machinescited by 144
  4. 2016Harnessing Deep Neural Networks with Logic Rulescited by 135
  5. 2022Semantic Probabilistic Layers for Neuro-Symbolic Learningcited by 103
  6. 2020Analyzing Differentiable Fuzzy Logic Operatorscited by 100
  7. 2023NeurASP: Embracing Neural Networks into Answer Set Programmingcited by 92
  8. 2020Inductive logic programming at 30: a new introductioncited by 83
  9. 2021From Statistical Relational to Neurosymbolic Artificial Intelligence: a Surveycited by 72
  10. 2015Hinge-Loss Markov Random Fields and Probabilistic Soft Logiccited by 63
  11. 2020Learning Reasoning Strategies in End-to-End Differentiable Provingcited by 60
  12. 2016Efficiency of minimizing compositions of convex functions and smooth mapscited by 59
  13. 2020Learning from Rules Generalizing Labeled Exemplarscited by 59
  14. 2020Closed Loop Neural-Symbolic Learning via Integrating Neural Perception, Grammar Parsing, and Symbolic Reasoningcited by 58
  15. 2019Augmenting Neural Networks with First-order Logiccited by 51
  16. 2019Embedding Symbolic Knowledge into Deep Networkscited by 49
  17. 2021Faster Single-loop Algorithms for Minimax Optimization without Strong Concavitycited by 47
  18. 2021MultiplexNet: Towards Fully Satisfied Logical Constraints in Neural Networkscited by 44
  19. 2022Deep Learning with Logical Constraintscited by 44
  20. 2019Competitive Gradient Descentcited by 32
  21. 2018Stochastic subgradient method converges at the rate O(k-1/4) on weakly convex functionscited by 31
  22. 2023Injecting Logical Constraints into Neural Networks via Straight-Through Estimatorscited by 12
  23. 2020Understanding Deep Architectures with Reasoning Layercited by 6
  24. 2019On the relation between Loss Functions and T-Normscited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 3; 2025: 4; 2026: 3.

10 papers.

  1. 2024A Neuro-Symbolic Benchmark Suite for Concept Quality and Reasoning Shortcutscited by 19
  2. 2024On the Independence Assumption in Neurosymbolic Learningcited by 18
  3. 2025Shortcuts and Identifiability in Concept-based Models from a Neuro-Symbolic Lenscited by 13
  4. 2025A Learnability Analysis on Neuro-Symbolic Learningcited by 3
  5. 2025The Road to Generalizable Neuro-Symbolic Learning Should be Paved with Foundation Modelscited by 3
  6. 2024MARS: A neurosymbolic approach for interpretable drug discoverycited by 2
  7. 2026Logic Tensor Network-Enhanced Generative Adversarial Networkcited by 2
  8. 2025Curriculum Abductive Learningcited by 1
  9. 2026A neurosymbolic Approach with Epistemic Deep Learning for Hierarchical Image Classificationcited by 1
  10. 2026Differentiable Logic Programming to Mitigate Reasoning Shortcuts in Neurosymbolic Systemscited by 0