metadata

annotations_creators:
  - expert-generated
language:
  - en
license: cc-by-nc-4.0
multilinguality: monolingual
pretty_name: CaseReportBench - Clinical Dense Extraction Benchmark
tags:
  - clinical-nlp
  - dense-information-extraction
  - medical
  - case-reports
  - rare-diseases
  - benchmarking
  - information-extraction
task_categories:
  - information-extraction
  - text-classification
  - question-answering
task_ids:
  - entity-extraction
  - multi-label-classification
  - open-domain-qa

CaseReportBench: Clinical Dense Extraction Benchmark

CaseReportBench is a curated benchmark dataset designed to evaluate the ability of large language models to perform dense information extraction from clinical case reports, particularly in the context of rare disease diagnosis.

This dataset supports fine-grained, system-wise phenotype extraction and structured diagnostic reasoning evaluation.

Key Features

Expert-annotated dense labels simulating comprehensive head-to-toe clinical assessments, capturing multi-system findings as encountered in real-world diagnostic reasoning
Domain: Clinical Case Reports (PubmedCentral indexed)
Use case: Medical IE, LLM evaluation, Rare disease diagnosis
Data type: JSON with structured system-wise output
Evaluation metrics: Token Selection Rate, Levenshtein Similarity, Exact Match

Dataset Structure

Each record includes:

id: Unique document identifier
text: Raw case report
extracted_labels: Dense structured annotations by system (e.g., nervous system, metabolic)
diagnosis: Gold standard diagnosis
source: PubMed ID or citation

Usage

from datasets import load_dataset

ds = load_dataset("cxyzhang/caseReportBench_ClinicalDenseExtraction_Benchmark")
print(ds["train"][0])

Citation

@inproceedings{zhang2025casereportbench,
  title={CaseReportBench: An LLM Benchmark Dataset for Dense Information Extraction in Clinical Case Reports},
  author={Zhang, Cindy and Others},
  booktitle={Conference on Health, Inference, and Learning (CHIL)},
  year={2025}
}