# Towards Automatic Evaluation and Selection of PHI De-identification Models via Multi-Agent Collaboration

arXiv: https://arxiv.org/abs/2510.16194 · HTML: https://mlsanity.com/paper/2510.16194/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Medical and scientific domain agents, one of 16 directions in LLM agents and tool use.
- Builds on 16 of its 32 references that are in the map.
- First posted on arXiv 17 Oct 2025.

## Placed in

- LLM agents and tool use / [Medical and scientific domain agents](https://mlsanity.com/agents/medical-and-scientific-domain-agents/)

## Builds on (most cited first, 16 of 16)

- [BioBERT: a pre-trained biomedical language representation model for biomedical text mining](https://mlsanity.com/paper/1901.08746/) (2019)
- [A Survey on Evaluation of Large Language Models](https://mlsanity.com/paper/2307.03109/) (2023)
- [Large Language Models are not Fair Evaluators](https://mlsanity.com/paper/2305.17926/) (2023)
- [Prometheus: Inducing Fine-grained Evaluation Capability in Language Models](https://mlsanity.com/paper/2310.08491/) (2023)
- [JudgeLM: Fine-tuned Large Language Models are Scalable Judges](https://mlsanity.com/paper/2310.17631/) (2023)
- [DeID-GPT: Zero-shot Medical Text De-Identification by GPT-4](https://mlsanity.com/paper/2303.11032/) (2023)
- [Do We Still Need Clinical Language Models?](https://mlsanity.com/paper/2302.08091/) (2023)
- [De-identification of Patient Notes with Recurrent Neural Networks](https://mlsanity.com/paper/1606.03475/) (2016)
- [Are Large Language Model-based Evaluators the Solution to Scaling Up Multilingual Evaluation?](https://mlsanity.com/paper/2309.07462/) (2023)
- [PromptLink: Leveraging Large Language Models for Cross-Source Biomedical Concept Linking](https://mlsanity.com/paper/2405.07500/) (2024)
- [De-identification of clinical free text using natural language processing: A systematic review of current approaches](https://mlsanity.com/paper/2312.03736/) (2023)
- [KERAP: A Knowledge-Enhanced Reasoning Approach for Accurate Zero-shot Diagnosis Prediction Using Multi-agent LLMs](https://mlsanity.com/paper/2507.02773/) (2025)
- [TACCO: Task-guided Co-clustering of Clinical Concepts and Patient Visits for Disease Subtyping based on EHR Data](https://mlsanity.com/paper/2406.10061/) (2024)
- [GraphNarrator: Generating Textual Explanations for Graph Neural Networks](https://mlsanity.com/paper/2410.15268/) (2024)
- [De-identification of medical records using conditional random fields and long short-term memory networks](https://mlsanity.com/paper/1709.06901/) (2017)
- [Large Language Model Empowered Privacy-Protected Framework for PHI Annotation in Clinical Notes](https://mlsanity.com/paper/2504.18569/) (2025)

## Built on by (most cited first, 3 of 3)

- [Utilizing Large Language Models for Zero-Shot Medical Ontology Extension from Clinical Notes](https://mlsanity.com/paper/2511.16548/) (2025)
- [Enhanced Atrial Fibrillation Prediction in ESUS Patients with Hypergraph-based Pre-training](https://mlsanity.com/paper/2603.13297/) (2026)
- [Institution-Specific LLM Prompting Recovers PHI That De-identification Systems and Their Gold Standards Both Miss](https://mlsanity.com/paper/2608.17051/) (2026)

## Abstract (arXiv)

Protected health information (PHI) de-identification is critical for enabling the safe reuse of clinical notes, yet evaluating and comparing PHI de-identification models typically depends on costly, small-scale expert annotations. We present TEAM-PHI, a multi-agent evaluation and selection framework that uses large language models (LLMs) to automatically measure de-identification quality and select the best-performing model without heavy reliance on gold labels. TEAM-PHI deploys multiple Evaluation Agents, each independently judging the correctness of PHI extractions and outputting structured metrics. Their results are then consolidated through an LLM-based majority voting mechanism that integrates diverse evaluator perspectives into a single, stable, and reproducible ranking. Experiments on a real-world clinical note corpus demonstrate that TEAM-PHI produces consistent and accurate rankings: despite variation across individual evaluators, LLM-based voting reliably converges on the same top-performing systems. Further comparison with ground-truth annotations and human evaluation confirms that the framework's automated rankings closely match supervised evaluation. By combining independent evaluation agents with LLM majority voting, TEAM-PHI offers a practical, secure, and cost-effective solution for automatic evaluation and best-model selection in PHI de-identification, even when ground-truth labels are limited.