ECGBench: Benchmark for Multimodal LLM Comprehension of Electrocardiographic Images

ECGBench is a benchmark dataset introduced in the paper 'Teach Multimodal LLMs to Comprehend Electrocardiographic Images'. The dataset was created by PULSE-ECG and last updated on Hugging Face on October 28, 2024. It is designed to evaluate the ability of multimodal large language models to understand and interpret ECG images.

Use Cases

• Benchmarking model performance on ECG image comprehension tasks based on the dataset's stated purpose.
• Training multimodal LLMs to interpret medical imagery based on the associated ECGInstruct dataset.
• Evaluating the clinical reasoning capabilities of AI models based on electrocardiographic data.

Strengths

• Dataset is directly linked to a peer-reviewed arXiv paper (arXiv:2410.19008).
• Associated with a public GitHub repository and a pre-trained model (PULSE-7B).
• Last updated date is explicitly provided (2024-10-28).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

PULSE-ECG

Collection Method

Likely created for the research project described in the associated paper.

Time Range

null

Freshness

Last updated 2024-10-28 01:29:48; freshness should be verified.

Geography

null