IDEA: An Interpretable and Editable Decision-Making Framework for LLMs via Verbal-to-Numeric Calibration

2026-04-14

AI Summary

Large language models (LLMs) are often limited in high-stakes decision-making due to poor calibration, weak interpretability, and difficulty incorporating expert knowledge. This work proposes the IDEA framework, which distills LLM decision knowledge into an interpretable, parameterized model grounded in semantic factors. By jointly learning a language-to-numeric calibration mapping and decision parameters, IDEA enables precise human-AI collaborative editing with formal mathematical guarantees. The approach supports perfect factor exclusion, exact probability calibration, and surpasses the performance ceiling of conventional prompt engineering. Evaluated on five benchmark datasets using Qwen-3-32B, IDEA achieves an accuracy of 78.6%, significantly outperforming DeepSeek R1 (68.1%) and GPT-5.2 (77.9%), while simultaneously ensuring perfect calibration and controllable model editing.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: (Large) Language Models
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Large Language Models are increasingly deployed for decision-making, yet their adoption in high-stakes domains remains limited by miscalibrated probabilities, unfaithful explanations, and inability to incorporate expert knowledge precisely. We propose IDEA, a framework that extracts LLM decision knowledge into an interpretable parametric model over semantically meaningful factors. Through joint learning of verbal-to-numerical mappings and decision parameters via EM, correlated sampling that preserves factor dependencies, and direct parameter editing with mathematical guarantees, IDEA produces calibrated probabilities while enabling quantitative human-AI collaboration. Experiments across five datasets show IDEA with Qwen-3-32B (78.6%) outperforms DeepSeek R1 (68.1%) and GPT-5.2 (77.9%), achieving perfect factor exclusion and exact calibration -- precision unattainable through prompting alone. The implementation is publicly available at https://github.com/leonbig/IDEA.

Problem

• calibration
• interpretability
• expert knowledge integration
• decision-making
• large language models

Innovation

• Interpretable AI
• Verbal-to-Numeric Calibration
• Parameter Editing
• EM Algorithm
• Calibrated Decision-Making