Towards Achieving Concept Completeness for Textual Concept Bottleneck Models
This addresses the challenge of interpretability in NLP classifiers by providing a fully unsupervised method to enhance concept completeness, though it appears incremental as it builds on existing TCBM frameworks.
The paper tackled the problem of achieving concept completeness in textual concept bottleneck models by proposing CT-CBM, which builds concept labels unsupervised using a small language model, eliminating the need for human or LLM annotations, and it achieved striking results in concept basis completeness and concept detection accuracy.
Textual Concept Bottleneck Models (TCBMs) are interpretable-by-design models for text classification that predict a set of salient concepts before making the final prediction. This paper proposes Complete Textual Concept Bottleneck Model (CT-CBM), a novel TCBM generator building concept labels in a fully unsupervised manner using a small language model, eliminating both the need for predefined human labeled concepts and LLM annotations. CT-CBM iteratively targets and adds important and identifiable concepts in the bottleneck layer to create a complete concept basis. CT-CBM achieves striking results against competitors in terms of concept basis completeness and concept detection accuracy, offering a promising solution to reliably enhance interpretability of NLP classifiers.