Datasets:

Saint-lsy
/

EndoBench

license: cc-by-sa-3.0
tags:
  - medical
  - endoscopy
  - VQA
  - VLM
language:
  - en
task_categories:
  - question-answering
configs:
  - config_name: EndoBench
    data_files:
      - split: test
        path: EndoBench.tsv
pretty_name: EndoBench
size_categories:
  - 1K<n<10K

EndoBench

🍎 Homepage|💻 GitHub|🤗 Dataset|📖 Paper

This repository is the official implementation of the paper EndoBench: A Comprehensive Evaluation of Multi-Modal Large Language Models for Endoscopy Analysis.

🚀 News

[21/10/2025] We release a new open-set challenging VQA benchmark EndoBench-extended.
[19/09/2025] 🎉🎉Our EndoBench was accepted by NeurIPS'25 D&B Track!!!

☀️ Tutorial

EndoBench is a comprehensive MLLM evaluation framework spanning 4 endoscopy scenarios and 12 clinical tasks with 12 secondary subtasks that mirror the progression of endoscopic examination workflows. Featuring five levels of visual prompting granularity to assess region-specific understanding, our EndoBench contains 6,832 clinically validated VQA pairs derived from 22 endoscopy datasets. This structure enables precise measurement of MLLMs' clinical perceptual, diagnostic accuracy, and spatial comprehension across diverse endoscopic scenarios.

Our dataset construction involves collecting 20 public and 1 private endoscopy datasets and standardizing QA pairs, yielding 446,535 VQA pairs comprising our EndoVQA-Instruct dataset, the current largest endoscopic instruction-tuning collection. From EndoVQA-Instruct, we extract representative pairs that undergo rigorous clinical review, resulting in our final EndoBench of 6,832 clinically validated VQA pairs.

We split EndoVQA-Instruct and provide two datasets:

EndoVQA-Instruct-trainval, which included 439703 VQA pairs. We provide the .json file containing the original image paths. You can download these datasets according to your needs. The private WCE2025 dataset is available upon request.
EndoBench, resulting in 6,832 rigorously validated VQA pairs, and we provide 2 versions of EndoBench: EndoBench.json and EndoBench.tsv file. Each data entry in the EndoBench.json file corresponds to an image in the EndoBench-Images.zip file. The EndoBench.tsv file contains images in base64 format.

Evaluation

The evaluation can be built upon VLMEvalKit. To get started:

This project is built upon VLMEvalKit. To get started:

Visit the VLMEvalKit Quickstart Guide for installation instructions. or you can run the following command for a quick start:

git clone https://github.com/CUHK-AIM-Group/EndoBench.git
cd VLMEvalKit
pip install -e .

You can evaluate your model with the following command:

python run.py --data EndoBench --model Your_model_name

Demo: Qwen2.5-VL-7B-Instruct on EndoBench, Inference only

python run.py --data EndoBench --model Qwen2.5-VL-7B-Instruct --mode infer

You can find more details on the ImageMCQDataset Class.

🏥 EndoBench-Extended

This EndoBench-extended dataset is the extended version of Endobench, which includes 48 images and 91 open-set VQA pairs, carefully selected by medical experts to ensure clinical relevance and diagnostic difficulty. The EndoBench-extended specifically focuses on challenging and underrepresented cases, including rare pathologies, overlapping lesions, atypical anatomical presentations, and post-surgical endoscopic views.

Disclaimers

The guidelines for the annotators emphasized strict compliance with copyright and licensing rules from the initial data source, specifically avoiding materials from websites that forbid copying and redistribution. Should you encounter any data samples potentially breaching the copyright or licensing regulations of any site, we encourage you to contact us. Upon verification, such samples will be promptly removed.

Greatly appreciate to all the authors of these datasets for their contributions to the field of endoscopy analysis.