Additive Large Language Models for Semi-Structured Text

2025-11-14

AI Summary

In clinical text classification, the lack of interpretability in large language model (LLM) predictions impedes their trustworthy adoption in research and clinical practice. To address this, we propose CALM—a framework for interpretable classification of semi-structured clinical texts. CALM first decomposes patient records into semantically coherent, human-understandable blocks; it then models LLM predictions as additive combinations of contributions from these blocks, embedding interpretability directly into the forward pass. Its key innovation is the explicit construction of LLM predictions as a generalized additive model (GA²M), enabling both instance-level feature attribution and population-level risk curve visualization. Experiments demonstrate that CALM preserves LLM-level classification accuracy while substantially enhancing model transparency, clinical auditability, and stakeholder trust—facilitating quality assurance and discovery of clinically meaningful patterns.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Computer Vision: Interpretability, Explainability, and Transparency

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
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Large Language Models have advanced clinical text classification, but their opaque predictions remain a critical barrier to practical adoption in research and clinical settings where investigators and physicians need to understand which parts of a patient's record drive risk signals. To address this challenge, we introduce extbf{CALM}, short for extbf{Classification with Additive Large Language Models}, an interpretable framework for semi-structured text where inputs are composed of semantically meaningful components, such as sections of an admission note or question-answer fields from an intake form. CALM predicts outcomes as the additive sum of each component's contribution, making these contributions part of the forward computation itself and enabling faithful explanations at both the patient and population level. The additive structure also enables clear visualizations, such as component-level risk curves similar to those used in generalized additive models, making the learned relationships easier to inspect and communicate. Although CALM expects semi-structured inputs, many clinical documents already have this form, and similar structure can often be automatically extracted from free-text notes. CALM achieves performance comparable to conventional LLM classifiers while improving trust, supporting quality-assurance checks, and revealing clinically meaningful patterns during model development and auditing.

Problem

• LLM predictions lack interpretability for clinical adoption
• Models need to identify driving factors in patient records
• Require faithful explanations at patient and population levels

Innovation

• Additive component contributions for interpretable predictions
• Faithful explanations at patient and population levels
• Component-level risk curves for clear visualizations