Prototype-Grounded Concept Models for Verifiable Concept Alignment

Stefano Colamonaco, David Debot, Pietro Barbiero, Giuseppe Marra
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:21212-21229, 2026.

Abstract

Concept Bottleneck Models (CBMs) aim to improve interpretability in Deep Learning by structuring predictions through human-understandable concepts, but they provide no way to verify whether learned concepts align with the human’s intended meaning, hurting interpretability. We introduce Prototype-Grounded Concept Models (PGCMs), which ground concepts in learned visual prototypes: image parts that serve as explicit evidence for the concepts. This grounding enables direct inspection of concept semantics and supports targeted human intervention at the prototype level to correct misalignments. Empirically, PGCMs achieve similar predictive performance as state-of-the-art CBMs while substantially improving transparency, interpretability, and intervenability.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-colamonaco26a, title = {Prototype-Grounded Concept Models for Verifiable Concept Alignment}, author = {Colamonaco, Stefano and Debot, David and Barbiero, Pietro and Marra, Giuseppe}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {21212--21229}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/colamonaco26a/colamonaco26a.pdf}, url = {https://proceedings.mlr.press/v306/colamonaco26a.html}, abstract = {Concept Bottleneck Models (CBMs) aim to improve interpretability in Deep Learning by structuring predictions through human-understandable concepts, but they provide no way to verify whether learned concepts align with the human’s intended meaning, hurting interpretability. We introduce Prototype-Grounded Concept Models (PGCMs), which ground concepts in learned visual prototypes: image parts that serve as explicit evidence for the concepts. This grounding enables direct inspection of concept semantics and supports targeted human intervention at the prototype level to correct misalignments. Empirically, PGCMs achieve similar predictive performance as state-of-the-art CBMs while substantially improving transparency, interpretability, and intervenability.} }
Endnote
%0 Conference Paper %T Prototype-Grounded Concept Models for Verifiable Concept Alignment %A Stefano Colamonaco %A David Debot %A Pietro Barbiero %A Giuseppe Marra %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-colamonaco26a %I PMLR %P 21212--21229 %U https://proceedings.mlr.press/v306/colamonaco26a.html %V 306 %X Concept Bottleneck Models (CBMs) aim to improve interpretability in Deep Learning by structuring predictions through human-understandable concepts, but they provide no way to verify whether learned concepts align with the human’s intended meaning, hurting interpretability. We introduce Prototype-Grounded Concept Models (PGCMs), which ground concepts in learned visual prototypes: image parts that serve as explicit evidence for the concepts. This grounding enables direct inspection of concept semantics and supports targeted human intervention at the prototype level to correct misalignments. Empirically, PGCMs achieve similar predictive performance as state-of-the-art CBMs while substantially improving transparency, interpretability, and intervenability.
APA
Colamonaco, S., Debot, D., Barbiero, P. & Marra, G.. (2026). Prototype-Grounded Concept Models for Verifiable Concept Alignment. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:21212-21229 Available from https://proceedings.mlr.press/v306/colamonaco26a.html.

Related Material