让大模型决策可解释可编辑，还能精准校准概率。
IDEA: An Interpretable and Editable Decision-Making Framework for LLMs via Verbal-to-Numeric Calibration
- 将语言决策转化为可调参数的数学模型，实现语义因子建模。
- 在5个数据集上准确率78.6%，实现完美因子排除与精确校准。
- 适合需要可信决策、可干预结果的高风险场景使用。
大语言模型在决策任务中日益应用，但其在高风险领域推广受限于概率校准不足、解释不忠实及难以精确融入专家知识。本文提出IDEA框架，通过联合学习语言到数值的映射与决策参数，结合保持因子相关性的采样方法，以及具有数学保证的参数直接编辑，将大模型决策知识提取为可解释的参数化模型。实验在五个数据集上显示，IDEA使用Qwen-3-32B模型达到78.6%准确率，优于DeepSeek R1（68.1%）和GPT-5.2（77.9%），实现完全因子排除与精确校准，这是仅靠提示无法达成的。代码已开源：https://github.com/leonbig/IDEA。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。