Shortcuts and Identifiability in Concept-based Models from a Neuro-Symbolic Lens
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866909990041681920 |
|---|---|
| author | Bortolotti, Samuele Marconato, Emanuele Morettin, Paolo Passerini, Andrea Teso, Stefano |
| author_facet | Bortolotti, Samuele Marconato, Emanuele Morettin, Paolo Passerini, Andrea Teso, Stefano |
| contents | Concept-based Models are neural networks that learn a concept extractor to map inputs to high-level concepts and an inference layer to translate these into predictions. Ensuring these modules produce interpretable concepts and behave reliably in out-of-distribution is crucial, yet the conditions for achieving this remain unclear. We study this problem by establishing a novel connection between Concept-based Models and reasoning shortcuts (RSs), a common issue where models achieve high accuracy by learning low-quality concepts, even when the inference layer is fixed and provided upfront. Specifically, we extend RSs to the more complex setting of Concept-based Models and derive theoretical conditions for identifying both the concepts and the inference layer. Our empirical results highlight the impact of RSs and show that existing methods, even combined with multiple natural mitigation strategies, often fail to meet these conditions in practice. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_11245 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Shortcuts and Identifiability in Concept-based Models from a Neuro-Symbolic Lens Bortolotti, Samuele Marconato, Emanuele Morettin, Paolo Passerini, Andrea Teso, Stefano Machine Learning Artificial Intelligence Concept-based Models are neural networks that learn a concept extractor to map inputs to high-level concepts and an inference layer to translate these into predictions. Ensuring these modules produce interpretable concepts and behave reliably in out-of-distribution is crucial, yet the conditions for achieving this remain unclear. We study this problem by establishing a novel connection between Concept-based Models and reasoning shortcuts (RSs), a common issue where models achieve high accuracy by learning low-quality concepts, even when the inference layer is fixed and provided upfront. Specifically, we extend RSs to the more complex setting of Concept-based Models and derive theoretical conditions for identifying both the concepts and the inference layer. Our empirical results highlight the impact of RSs and show that existing methods, even combined with multiple natural mitigation strategies, often fail to meet these conditions in practice. |
| title | Shortcuts and Identifiability in Concept-based Models from a Neuro-Symbolic Lens |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2502.11245 |