Concept Bottleneck Large Language Models

2024-12-11

arXiv.org

AI Summary

Large language models (LLMs) suffer from poor interpretability, low safety, and limited trustworthiness due to their inherent “black-box” nature. Method: This paper proposes an intrinsically interpretable LLM construction paradigm, introducing the first concept-bottleneck architecture for LLMs. It incorporates concept-constrained fine-tuning, interpretable neuron localization, concept-space projection, and intervention techniques into Transformer-based models to explicitly encode human-understandable semantic concepts. This enables concept-level detection and generation guidance across diverse tasks. Contribution/Results: The approach achieves text classification performance competitive with state-of-the-art black-box models while providing precise, faithful attributions. For text generation, it supports fine-grained, concept-driven control—substantially improving explanation quality and human-model collaboration. To our knowledge, this is the first work to realize an embedded, intervenable, and general-purpose interpretability mechanism within LLMs.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Computer Vision: Large Vision Models

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

We introduce the Concept Bottleneck Large Language Model (CB-LLM), a pioneering approach to creating inherently interpretable Large Language Models (LLMs). Unlike traditional black-box LLMs that rely on post-hoc interpretation methods with limited neuron function insights, CB-LLM sets a new standard with its built-in interpretability, scalability, and ability to provide clear, accurate explanations. We investigate two essential tasks in the NLP domain: text classification and text generation. In text classification, CB-LLM narrows the performance gap with traditional black-box models and provides clear interpretability. In text generation, we show how interpretable neurons in CB-LLM can be used for concept detection and steering text generation. Our CB-LLMs enable greater interaction between humans and LLMs across a variety of tasks -- a feature notably absent in existing LLMs. Our code is available at https://github.com/Trustworthy-ML-Lab/CB-LLMs.

Problem

• Enhancing interpretability in Large Language Models (LLMs).
• Improving safety and reliability in text generation tasks.
• Providing transparent and scalable explanations for NLP tasks.

Innovation

• CB-LLMs integrate intrinsic interpretability directly
• CB-LLMs enable precise concept detection in generation
• Embedded interpretability enhances safety and reliability