arXiv 2601.119052026-01-17cs.AIcs.LGmath.STstat.TH

LIBRA：基于语言模型的带状 recourse 算法用于个性化治疗计划

LIBRA: Language Model Informed Bandit Recourse Algorithm for Personalized Treatment Planning

McCombs School of Business, University of Texas at Austin（德克萨斯大学奥斯汀分校麦克斯韦商学院）；Naveen Jindal School of Management, University of Texas at Dallas（德克萨斯大学达拉斯分校奈文·金达管理学院）；Mays Business School, Texas A&M University（德克萨斯农工大学梅斯商学院）；Wharton School, University of Pennsylvania（宾夕法尼亚大学沃顿商学院）

Junyu Cao, Ruijiang Gao, Esmaeil Keyvanshokooh, Jianhao Ma

备注：: 50 pages. Previous version with human-AI collaboration: arXiv:2410.14640

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

LIBRA 是一种结合大语言模型和带状学习的算法，用于在个性化治疗中实现更高效的决策和鲁棒性。

摘要：

我们介绍了一个统一的框架，该框架无缝整合了算法性 recourse、上下文带状和大语言模型（LLMs），以支持在高风险领域如个性化医学中的顺序决策。我们首先介绍了 recourse 带状问题，其中决策者必须选择治疗行动和可行的、最小的可变患者特征修改。为了解决这个问题，我们开发了广义线性 recourse 带状（GLRB）算法。在此基础上，我们提出 LIBRA，一种语言模型指导的带状 recourse 算法，战略性地结合了来自 LLMs 的领域知识与带状学习的统计严谨性。LIBRA 提供了三个关键保证：（i）预热保证，显示 LIBRA 在 LLM 建议接近最优时显著减少初始遗憾；（ii）LLM 努力保证，证明该算法仅在 O(log² T) 次咨询 LLM，其中 T 是时间范围，确保长期自主性；（iii）鲁棒性保证，显示即使 LLM 不可靠，LIBRA 也永远不会比纯带状算法表现更差。我们进一步建立了匹配的下界，这些下界描述了 recourse 带状问题的根本难度，并展示了我们算法的近最优性。在合成环境和一个真实的高血压管理案例研究中，GLRB 和 LIBRA 与标准上下文带状和 LLM 仅基准相比，提高了遗憾、治疗质量和样本效率。我们的结果突显了 recourse 意识、LLM 协助带状算法在个性化高风险决策中的潜力。

英文摘要：

We introduce a unified framework that seamlessly integrates algorithmic recourse, contextual bandits, and large language models (LLMs) to support sequential decision-making in high-stakes settings such as personalized medicine. We first introduce the recourse bandit problem, where a decision-maker must select both a treatment action and a feasible, minimal modification to mutable patient features. To address this problem, we develop the Generalized Linear Recourse Bandit (GLRB) algorithm. Building on this foundation, we propose LIBRA, a Language Model-Informed Bandit Recourse Algorithm that strategically combines domain knowledge from LLMs with the statistical rigor of bandit learning. LIBRA offers three key guarantees: (i) a warm-start guarantee, showing that LIBRA significantly reduces initial regret when LLM recommendations are near-optimal; (ii) an LLM-effort guarantee, proving that the algorithm consults the LLM only $O(\log^2 T)$ times, where $T$ is the time horizon, ensuring long-term autonomy; and (iii) a robustness guarantee, showing that LIBRA never performs worse than a pure bandit algorithm even when the LLM is unreliable. We further establish matching lower bounds that characterize the fundamental difficulty of the recourse bandit problem and demonstrate the near-optimality of our algorithms. Experiments on synthetic environments and a real hypertension-management case study confirm that GLRB and LIBRA improve regret, treatment quality, and sample efficiency compared with standard contextual bandits and LLM-only benchmarks. Our results highlight the promise of recourse-aware, LLM-assisted bandit algorithms for trustworthy LLM-bandits collaboration in personalized high-stakes decision-making.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。