Mixture of Concept Bottleneck Experts
Francesco De Santis, Gabriele Ciravegna, Giovanni De Felice, Arianna Casanova, Francesco Giannini, Michelangelo Diligenti, Mateo Espinosa Zarlenga, Pietro Barbiero, Johannes Schneider, Danilo Giordano · 2026
Read Mixture of Concept Bottleneck Experts by Francesco De Santis, Gabriele Ciravegna, Giovanni De Felice, Arianna Casanova, Francesco Giannini, Michelangelo Diligenti, Mateo Espinosa Zarlenga, Pietro Barbiero, Johannes Schneider, Danilo Giordano (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-02886.txt.
Description
Francesco De Santis and 9 co-authors work on making model predictions easier to explain in a 2026 paper. Concept bottleneck models force predictions through human-understandable concepts first. Most such models then use one fixed formula to turn concepts into a final answer, which limits accuracy and flexibility.
Their framework, M-CBE, allows several formulas, called experts, instead of one. Two versions are tested. One learns a set of straight-line formulas. The other discovers formulas by symbolic regression within vocabularies the user chooses. Experiments show that changing the number and form of experts helps balance accuracy against interpretability. Different users can pick different points on that trade-off.
Catalog record
| CREATOR | Francesco De Santis, Gabriele Ciravegna, Giovanni De Felice, Arianna Casanova, Francesco Giannini, Michelangelo Diligenti, Mateo Espinosa Zarlenga, Pietro Barbiero, Johannes Schneider, Danilo Giordano |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 107 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-02886