[2510.14538] Symbol Grounding in Neuro-Symbolic AI: A Gentle Introduction to Reasoning Shortcuts

Computer Science > Artificial Intelligence arXiv:2510.14538 (cs) [Submitted on 16 Oct 2025 (v1), last revised 31 Mar 2026 (this version, v2)] Title:Symbol Grounding in Neuro-Symbolic AI: A Gentle Introduction to Reasoning Shortcuts Authors:Emanuele Marconato, Samuele Bortolotti, Emile van Krieken, Paolo Morettin, Elena Umili, Antonio Vergari, Efthymia Tsamoura, Andrea Passerini, Stefano Teso View a PDF of the paper titled Symbol Grounding in Neuro-Symbolic AI: A Gentle Introduction to Reasoning Shortcuts, by Emanuele Marconato and 8 other authors View PDF HTML (experimental) Abstract:Neuro-symbolic (NeSy) AI aims to develop deep neural networks whose predictions comply with prior knowledge encoding, e.g. safety or structural constraints. As such, it represents one of the most promising avenues for reliable and trustworthy AI. The core idea behind NeSy AI is to combine neural and symbolic steps: neural networks are typically responsible for mapping low-level inputs into high-level symbolic concepts, while symbolic reasoning infers predictions compatible with the extracted concepts and the prior knowledge. Despite their promise, it was recently shown that - whenever the concepts are not supervised directly - NeSy models can be affected by Reasoning Shortcuts (RSs). That is, they can achieve high label accuracy by grounding the concepts incorrectly. RSs can compromise the interpretability of the model's explanations, performance in out-of-distribution scenarios, and therefore...