Towards Achieving Concept Completeness for Unsupervised Textual Concept Bottleneck Models

ACL ARR 2025 February Submission4603 Authors

Abstract:

Textual Concept Bottleneck Models (TBMs) 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 approach generating concept labels in a fully unsupervised manner using a small language model, eliminating the need for predefined labeled concepts and the use of LLM for concept annotation. CT-CBM iteratively targets and adds important concepts in the bottleneck layer to create a nearly complete concept basis and addresses downstream classification leakage through a parallel residual connection.
CT-CBM achieves good results against competitors, offering a promising solution to enhance interpretability of NLP classifiers without sacrificing performance.

Paper Type: Long

Research Area: Interpretability and Analysis of Models for NLP

Research Area Keywords: Interpretability, Explainable AI, Concepts, Concepts Bottleneck Models

Contribution Types: Model analysis & interpretability

Languages Studied: English

Submission Number: 4603