On the Caltech-UCSD Birds dataset, post-hoc concept bottleneck models frequently rank concepts that are absent from the image or localize present concepts to the wrong region, as measured by the new CGIM, CEM, and CLM metrics.
best guess
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Concept-Based Explainable Artificial Intelligence: Metrics and Benchmarks
On the Caltech-UCSD Birds dataset, post-hoc concept bottleneck models frequently rank concepts that are absent from the image or localize present concepts to the wrong region, as measured by the new CGIM, CEM, and CLM metrics.