base_model: unsloth/Qwen3-30B-A3B-Instruct
library_name: peft
pipeline_tag: text-generation
tags:
- lora
- sft
- transformers
- unsloth
- biomedical
- phenotype-recognition
AutoPCR: Automated Phenotype Concept Recognition by Prompting
AutoPCR is a prompt-based phenotype concept recognition (CR) method designed to automatically generalize to new ontologies and unseen data without ontology-specific training. This repository contains the fine-tuned entity linker component of the system, which is a LoRA adapter for unsloth/Qwen3-30B-A3B-Instruct.
- Repository: https://github.com/yctao7/AutoPCR
- Paper: AutoPCR: Automated Phenotype Concept Recognition by Prompting
Model Description
Phenotype concept recognition (CR) is a fundamental task in biomedical text mining. Existing methods often struggle to generalize across diverse text styles or require extensive ontology-specific training. AutoPCR addresses these limitations by using a prompt-based approach and an optional self-supervised training strategy to achieve robust performance across multiple datasets. This model specifically serves as the entity linker within the pipeline to map extracted phenotype mentions to standard ontologies like HPO and MEDIC.
Usage
For detailed instructions on how to use this model within the AutoPCR framework—including environment setup, dictionary building, indexing, and running evaluation experiments—please refer to the official GitHub repository.
Example command for running HPO evaluation from the source code:
python HPO_evaluation.py --ontology_dict ../dict/HPO -c BIOC-GS -o ../results/bioc-gs.tsv --only_longest
Citation
If you find this work useful, please cite:
BibTeX:
@misc{tao2025autopcr,
title={AutoPCR: Automated Phenotype Concept Recognition by Prompting},
author={Yichao Tao and others},
year={2025},
eprint={2507.19315},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2507.19315},
}
Contact
Contact: drjieliu@umich.edu