Leaderboard
The Performance of Different VLMs on Neuro-Medical Tasks
Direct Diagnosis
Factual Medical Knowledge Application in Diagnostic Scenarios.
80
sample size
SOTA: 46.7%
Complex Diseases
Diagnostic Challenges in Complex and Rare Diseases.
20
sample size
SOTA: 40.0%
Multi-round Dialogue
Contextual Relationships in Multi-Round Dialogue.
100
sample size
SOTA: 18.5%
NeuralMedBench 2.0 Leaderboard
On the way here…
The forthcoming Leaderboard v2.0 is currently under active development and will feature an expanded set of tasks and metrics tailored for neuromedical multimodal reasoning. Coming soon!
Overview of Neural-MedBench
Dataset
We offer carefully curated and annotated neuromedicine datasets designed to support the training and evaluation of VLM models.
NeuralMedBench 2.0
On the way here…
The forthcoming Dataset v2.0 is currently under active development and will incorporate a broader collection of high-quality neuromedical datasets. It is designed to support increasingly complex multimodal tasks and to facilitate more accurate diagnostic evaluations. Coming soon!
News
Latest updates related to Neural-MedBench
2025-09-17 Neural-Benchmark: Pioneering Research Now Live on arXiv.
2025-09-12 Leaderboard is on! Check out the result!
2025-05-17 We release the evaluation code! Check out the Usage.
2025-05-16 We release the Neural-MedBench dataset
We introduce Neural-MedBench
A compact yet reasoning-intensive benchmark specifically designed to probe the limits of multimodal clinical reasoning in neurology.
About Us
Pioneering the exploration of large models in neuro-medicine, empowering AI-driven precision diagnosis and treatment of neurological disorders.
Hugging Face
Citation
DOI: 10.xxxx/xxxx
Additional citation details can go here. This box supports longer text with scroll.
Disclaimer
Neural-MedBench is for research purposes only. Models evaluated on Neural-MedBench can produce unexpected results. We are not responsible for any damages caused by the use of Neural-MedBench, including but not limited to, any loss of profit, data, or use of data.
License
This project is licensed under the MIT License.
Official Website
Follow Us
