# Surrogate modeling for interpreting black-box LLMs in medical predictions

arXiv: https://arxiv.org/abs/2604.20331 · HTML: https://mlsanity.com/paper/2604.20331/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 7 of its 75 references that are in the map.
- First posted on arXiv 22 Apr 2026.

## Builds on (most cited first, 7 of 7)

- [A Unified Approach to Interpreting Model Predictions](https://mlsanity.com/paper/1705.07874/) (2017)
- [Large Language Models Encode Clinical Knowledge](https://mlsanity.com/paper/2212.13138/) (2022)
- [GatorTron: A Large Clinical Language Model to Unlock Patient Information from Unstructured Electronic Health Records](https://mlsanity.com/paper/2203.03540/) (2022)
- [MedCoT: Medical Chain of Thought via Hierarchical Expert](https://mlsanity.com/paper/2412.13736/) (2024)
- [Large Language Models are Clinical Reasoners: Reasoning-Aware Diagnosis Framework with Prompt-Generated Rationales](https://mlsanity.com/paper/2312.07399/) (2023)
- [Perspectives on the Social Impacts of Reinforcement Learning with Human Feedback](https://mlsanity.com/paper/2303.02891/) (2023)
- [Cross-Care: Assessing the Healthcare Implications of Pre-training Data on Language Model Bias](https://mlsanity.com/paper/2405.05506/) (2024)

## Abstract (arXiv)

Large language models (LLMs), trained on vast datasets, encode extensive real-world knowledge within their parameters, yet their black-box nature obscures the mechanisms and extent of this encoding. Surrogate modeling, which uses simplified models to approximate complex systems, can offer a path toward better interpretability of black-box models. We propose a surrogate modeling framework that quantitatively explains LLM-encoded knowledge. For a specific hypothesis derived from domain knowledge, this framework approximates the latent LLM knowledge space using observable elements (input-output pairs) through extensive prompting across a comprehensive range of simulated scenarios. Through proof-of-concept experiments in medical predictions, we demonstrate our framework's effectiveness in revealing the extent to which LLMs "perceive" each input variable in relation to the output. Particularly, given concerns that LLMs may perpetuate inaccuracies and societal biases embedded in their training data, our experiments using this framework quantitatively revealed both associations that contradict established medical knowledge and the persistence of scientifically refuted racial assumptions within LLM-encoded knowledge. By disclosing these issues, our framework can act as a red-flag indicator to support the safe and reliable application of these models.