Mitigating Neuro-Symbolic Reasoning Shortcuts with Data-Driven Knowledge Augmentation
Open Access
- 8 August 2026
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 2861-2872
Abstract
Recent advancements in neuro-symbolic learning (NeSy) have shown significant promise in integrating deep learning with symbolic reasoning, offering both interpretability and generalization. However, the prevalence of reasoning shortcuts, where the NeSy system predicts incorrect intermediate concepts while maintaining high final accuracy, poses a substantial challenge. This is especially problematic in domains requiring reliable and transparent decision-making. Inspired by recent theories, we find that existing methods fail to address the reasoning shortcut issue when the knowledge base lacks sufficient complexity, highlighting their vulnerability in real-world applications. In this work, we present a novel method called DKA to address this issue. It introduces a limited set of concept-supervised data to enhance the knowledge base, effectively solving the reasoning shortcut problem and improving the applicability of the NeSy system. Theoretical analysis reveals that DKA can reduce shortcut risks with improved data efficiency. Empirical studies across multiple tasks within various neuro-symbolic frameworks also verify the effectiveness of the DKA method.Keywords
This publication has 14 references indexed in Scilit:
- Markov conditions and factorization in logical credal networksInternational Journal of Approximate Reasoning, 2024
- From statistical relational to neurosymbolic artificial intelligence: A surveyArtificial Intelligence, 2024
- A Survey on Deep Semi-Supervised LearningIEEE Transactions on Knowledge and Data Engineering, 2022
- Interactive Disentanglement: Learning Concepts by Interacting with their Prototype Representations2022
- Concept Bottleneck Model With Additional Unsupervised ConceptsIEEE Access, 2022
- Abductive Learning with Ground Knowledge Base2021
- Semi-Supervised Abductive Learning and Its Application to Theft Judicial Sentencing2020
- From Statistical Relational to Neuro-Symbolic Artificial Intelligence2020
- Explainable Object-Induced Action Decision for Autonomous Vehicles2020
- Gradient-based learning applied to document recognitionProceedings of the IEEE, 1998