Constraint-Based Analysis of Reasoning Shortcuts in Neurosymbolic Learning

2026-04-25

AI Summary

This work addresses a critical limitation in neuro-symbolic systems: despite satisfying logical constraints, they may exploit reasoning shortcuts due to incorrect concept-to-label mappings. The paper formalizes this issue as a constraint satisfaction problem and characterizes the conditions under which constraints uniquely determine the intended concept mapping. It introduces the first sound and complete answer set programming (ASP)-based verification algorithm alongside a greedy repair strategy. Theoretical analysis establishes that verifying the absence of shortcuts is coNP-complete, counting valid mappings is #P-complete, and computing a minimal repair is NP-hard; under optimal conditions, the sample query complexity achieves logarithmic bounds. Empirical evaluation across eight benchmark domains demonstrates the efficacy of the proposed approach.

Technology Category

• Constraint Satisfaction and Optimization: Satisfiability
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning
• Machine Learning: Neuro-Symbolic Learning

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Neurosymbolic systems can satisfy logical constraints during learning without achieving the intended concept-label correspondence; this is a problem known as reasoning shortcuts. We formalize reasoning shortcuts as a constraint satisfaction problem and investigate under which conditions concept mappings are uniquely determined by the constraints. We prove that a discrimination property (requiring that no valid concept mapping can be transformed into another valid mapping by swapping two concept values) is necessary for shortcut-freeness under bijective mappings, but demonstrate via a counterexample that it is insufficient even when the constraint graph is connected. We develop an ASP-based algorithm that verifies whether a given constraint set uniquely determines the intended concept mapping, with proven soundness and completeness. When shortcuts are detected, a greedy repair algorithm eliminates them by augmenting the constraint set, converging in at most $k$ iterations, where $k$ is the number of alternative valid mappings. We further provide a complexity classification: deciding shortcut-freeness is coNP-complete, counting shortcuts is #P-complete, and finding minimal repairs is NP-hard. We also establish sample complexity bounds showing that logarithmically many label queries suffice for disambiguation in favorable cases, while querying all ambiguous positions suffices in the worst case. Experiments across eight benchmark domains validate our approach.

Problem

• reasoning shortcuts
• neurosymbolic learning
• constraint satisfaction
• concept-label correspondence
• shortcut-freeness

Innovation

• reasoning shortcuts
• neurosymbolic learning
• constraint satisfaction
• answer set programming (ASP)
• concept-label correspondence