Holtercare-Bench: A Multimodal Benchmark for Evaluating Long-Term Dynamic ECG Analysis
Not provided in the abstract
Abstract
The paper introduces Holtercare-23K, a large-scale multimodal dynamic ECG dataset and Holtercare-Bench, a benchmark for evaluating models on dynamic ECG analysis.
Reality Card
The introduction of Holtercare-23K and Holtercare-Bench significantly enhances the evaluation of multimodal large language models in the context of long-term dynamic ECG analysis.
Zero-shot evaluations reveal a significant performance gap in processing ultra-long pathological sequences, but fine-tuning improves results substantially.
Current multimodal large language models struggle with complex temporal reasoning and diagnostic report generation due to limited high-quality datasets.
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.