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

arXiv: https://arxiv.org/abs/2601.11905 · HTML: https://mlsanity.com/paper/2601.11905/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 26 of its 79 references that are in the map.
- First posted on arXiv 17 Jan 2026.

## Builds on (most cited first, 26 of 26)

- [Counterfactual Explanations without Opening the Black Box: Automated Decisions and the GDPR](https://mlsanity.com/paper/1711.00399/) (2017)
- [Introduction to Multi-Armed Bandits](https://mlsanity.com/paper/1904.07272/) (2019)
- [Actionable Recourse in Linear Classification](https://mlsanity.com/paper/1809.06514/) (2018)
- [Learning to Optimize Via Posterior Sampling](https://mlsanity.com/paper/1301.2609/) (2013)
- [Model-Agnostic Counterfactual Explanations for Consequential Decisions](https://mlsanity.com/paper/1905.11190/) (2019)
- [Multi-Objective Counterfactual Explanations](https://mlsanity.com/paper/2004.11165/) (2020)
- [Learning Model-Agnostic Counterfactual Explanations for Tabular Data](https://mlsanity.com/paper/1910.09398/) (2019)
- [Preserving Causal Constraints in Counterfactual Explanations for Machine Learning Classifiers](https://mlsanity.com/paper/1912.03277/) (2019)
- [Algorithmic recourse under imperfect causal knowledge: a probabilistic approach](https://mlsanity.com/paper/2006.06831/) (2020)
- [A survey of algorithmic recourse: definitions, formulations, solutions, and prospects](https://mlsanity.com/paper/2010.04050/) (2020)
- [Towards Robust and Reliable Algorithmic Recourse](https://mlsanity.com/paper/2102.13620/) (2021)
- [On the Adversarial Robustness of Causal Algorithmic Recourse](https://mlsanity.com/paper/2112.11313/) (2021)
- [Algorithmic Recourse in the Wild: Understanding the Impact of Data and Model Shifts](https://mlsanity.com/paper/2012.11788/) (2020)
- [Human-AI Collaboration with Bandit Feedback](https://mlsanity.com/paper/2105.10614/) (2021)
- [Conversational Contextual Bandit: Algorithm and Application](https://mlsanity.com/paper/1906.01219/) (2019)
- [Do LLM Agents Have Regret? A Case Study in Online Learning and Games](https://mlsanity.com/paper/2403.16843/) (2024)
- [Large Language Models for Market Research: A Data-augmentation Approach](https://mlsanity.com/paper/2412.19363/) (2024)
- [LOLA: LLM-Assisted Online Learning Algorithm for Content Experiments](https://mlsanity.com/paper/2406.02611/) (2024)
- [The Best Decisions Are Not the Best Advice: Making Adherence-Aware Recommendations](https://mlsanity.com/paper/2209.01874/) (2022)
- [Jump Starting Bandits with LLM-Generated Prior Knowledge](https://mlsanity.com/paper/2406.19317/) (2024)
- [A Bandit Model for Human-Machine Decision Making with Private Information and Opacity](https://mlsanity.com/paper/2007.04800/) (2020)
- [Hierarchical Conversational Preference Elicitation with Bandit Feedback](https://mlsanity.com/paper/2209.06129/) (2022)
- [Leveraging Initial Hints for Free in Stochastic Linear Bandits](https://mlsanity.com/paper/2203.04274/) (2022)
- [Confounding-Robust Policy Improvement with Human-AI Teams](https://mlsanity.com/paper/2310.08824/) (2023)
- [Contextual Bandits with Budgeted Information Reveal](https://mlsanity.com/paper/2305.18511/) (2023)
- [A note on polynomial solvability of the CDT problem](https://mlsanity.com/paper/1406.6429/) (2014)

## Abstract (arXiv)

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(log2 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.