CREAMs architecturally encode concept-concept and concept-task relationships in CBMs plus an optional regularized side-channel, achieving competitive performance and improved intervenability under incomplete concepts.
Upper bound of bayesian generalization error in partial concept bottleneck model (cbm): Partial cbm outperforms naive cbm
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Towards Reasonable Concept Bottleneck Models
CREAMs architecturally encode concept-concept and concept-task relationships in CBMs plus an optional regularized side-channel, achieving competitive performance and improved intervenability under incomplete concepts.