ECG-InterpBench: Benchmarking the Interpretability of ECG Foundation Models with Matched-Scale Sparse Autoencoders

• URL: http://arxiv.org/abs/2607.27404v1
• Date: Wed, 29 Jul 2026 19:20:50 GMT
• Title: ECG-InterpBench: Benchmarking the Interpretability of ECG Foundation Models with Matched-Scale Sparse Autoencoders
• Abstract summary: ECG-InterpBench is a benchmark designed to evaluate the interpretability of ECG foundation-model representations.<n>We evaluate six frozen ECG foundation models across five standardized depths, five matched dictionary widths, and three random seeds.<n>The benchmark reveals that ECG foundation models exhibit distinct interpretability profiles.
• Score: 0.8162386935076237
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Existing benchmarks for electrocardiogram foundation models primarily evaluate downstream predictive performance, providing limited insight into whether their internal representations can be faithfully decomposed, clinically interpreted, or reproduced across independent analyses. We introduce ECG-InterpBench, a benchmark designed to systematically evaluate the interpretability of ECG foundation-model representations. ECG-InterpBench uses sparse autoencoders as standardized measurement instruments and matches their capacity across models to enable controlled comparisons. We evaluate six frozen ECG foundation models across five standardized encoder depths, five matched dictionary widths, and three random seeds, producing a 450-cell interpretability atlas comprising 75 exactly matched six-model comparison blocks. The benchmark evaluates complementary dimensions of representation interpretability, including sparse reconstruction fidelity, single-feature accessibility and coverage of 49 clinically meaningful ECG measurements, and cross-seed feature reproducibility. The evaluation further quantifies patient-sampling uncertainty, depth- and seed-dependent variation, and sensitivity to the sparsity parameterization. The benchmark reveals that ECG foundation models exhibit distinct interpretability profiles. A matched replication on MIMIC-IV-ECG confirms that reconstruction fidelity and clinical accessibility identify different leading models. The benchmark is accompanied by executable evaluation code, standardized manifests, cell-level metrics, and reproducibility audits. ECG-InterpBench complements performance-centered ECG benchmarks by providing a capacity-controlled and reproducible framework for comparing ECG foundation models across distinct dimensions of representation interpretability.

Related papers

• ProCA: Progressive Contrastive Alignment for Robust EEG Visual Decoding [47.35612571473169]
  EEG visual decoding aims to recover visual semantics from non-invasive neural time-series signals.<n>Existing methods rely on fixed visual or textual anchors whose semantic relations may become misaligned with EEG representations that vary across trials, subjects, and learning stages.<n>We propose Progressive Contrastive Alignment (ProCA), a unified and model-agnostic framework for adaptive neural-semantic alignment.
  arXiv  Detail & Related papers  (2026-09-04T12:48:25Z)
• FOUND-AF: Benchmarking ECG Foundation Models for Atrial Fibrillation Detection [5.65079592827682]
  Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia and is associated with increased risks of stroke, heart failure, and mortality.<n>Recent ECG foundation models offer transferable representations for automated AF detection.<n>This study presents FOUND-AF, a unified, leakage-controlled, and deployment-oriented benchmarking framework that evaluates the quality of pretrained ECG representations.
  arXiv  Detail & Related papers  (2026-08-04T12:48:05Z)
• Latent Performance Profiling of Large Language Models [47.009623327601226]
  We introduce Latent Performance Profiling (LPP) -- a framework that derives task-agnostic diagnostics from hidden activations and output distributions.<n>Unlike static accuracy scores, LPP provides stable, architecture-sensitive signatures across models of similar size.<n>We show that models with similar benchmark scores can exhibit contrasting latent profiles, such as differences in entropy or adaptability.
  arXiv  Detail & Related papers  (2026-05-28T14:41:26Z)
• CriterAlign: Criterion-Centric Rationale Alignment for Code Preference Judging [95.02210956333374]
  We propose a criterion-centric framework that adapts rubric-based judging to pairwise preference evaluation.<n>On BigCodeReward, CriterAlign improves a Qwen2.5-VL-32B monolithic judge from 60.4% to 66.3% accuracy.
  arXiv  Detail & Related papers  (2026-05-19T10:59:19Z)
• Detecting low left ventricular ejection fraction from ECG using an interpretable and scalable predictor-driven framework [9.789732284839884]
  Low left ventricular ejection fraction (LEF) frequently remains undetected until progression to symptomatic heart failure.<n>Existing approaches rely solely on end-to-end black-box models with limited interpretability.<n>We introduced ECG-based Predictor-Driven LEF (ECGPD-LEF), a structured framework that integrates foundation model-derived diagnostic probabilities with interpretable modeling.
  arXiv  Detail & Related papers  (2026-03-30T14:55:52Z)
• ECG-Reasoning-Benchmark: A Benchmark for Evaluating Clinical Reasoning Capabilities in ECG Interpretation [14.296134981531337]
  We introduce a novel multi-turn evaluation framework comprising over 6,400 samples to assess step-by-step reasoning across 17 core ECG diagnoses.<n>Our comprehensive evaluation of state-of-the-art models reveals a critical failure in executing multi-step logical deduction.<n>These results demonstrate that current MLLMs bypass actual visual interpretation, exposing a critical flaw in existing training paradigms and underscoring the necessity for robust, reasoning-centric medical AI.
  arXiv  Detail & Related papers  (2026-03-15T11:09:38Z)
• IF-RewardBench: Benchmarking Judge Models for Instruction-Following Evaluation [85.56193980646981]
  We propose IF-RewardBench, a comprehensive meta-evaluation benchmark for instruction-following.<n>For each instruction, we construct a preference graph containing all pairwise preferences among multiple responses.<n>Experiments on IF-RewardBench reveal significant deficiencies in current judge models.
  arXiv  Detail & Related papers  (2026-03-05T02:21:17Z)
• Position: Evaluation of ECG Representations Must Be Fixed [1.567009619451362]
  This position paper argues that current benchmarking practice in 12-lead ECG representation learning must be fixed to ensure progress is reliable and aligned with clinically meaningful objectives.<n>We argue that downstream evaluation should expand to include an assessment of structural heart disease and patient-level forecasting, in addition to other evolving ECG-related endpoints, as relevant clinical targets.
  arXiv  Detail & Related papers  (2026-02-19T16:42:46Z)
• ECG-IMN: Interpretable Mesomorphic Neural Networks for 12-Lead Electrocardiogram Interpretation [1.7152890789230053]
  Deep learning has achieved expert-level performance in automated electrocardiogram (ECG) diagnosis, yet the "black-box" nature of these models hinders their clinical deployment.<n>We propose the ECG-IMN, an Interpretable Mesomorphic Neural Network tailored for high-resolution 12-lead ECG classification.
  arXiv  Detail & Related papers  (2026-02-10T09:17:29Z)
• GEM: Empowering MLLM for Grounded ECG Understanding with Time Series and Images [44.50428701650495]
  We introduce GEM, the first MLLM unifying ECG time series, 12-lead ECG images and text for grounded and clinician-aligned ECG interpretation.<n> GEM enables feature-grounded analysis, evidence-driven reasoning, and a clinician-like diagnostic process through three core innovations.<n>We propose the Grounded ECG task, a clinically motivated benchmark designed to assess the MLLM's capability in grounded ECG understanding.
  arXiv  Detail & Related papers  (2025-03-08T05:48:53Z)
• ECG-DelNet: Delineation of Ambulatory Electrocardiograms with Mixed Quality Labeling Using Neural Networks [69.25956542388653]
  Deep learning (DL) algorithms are gaining weight in academic and industrial settings. We demonstrate DL can be successfully applied to low interpretative tasks by embedding ECG detection and delineation onto a segmentation framework. The model was trained using PhysioNet's QT database, comprised of 105 ambulatory ECG recordings.
  arXiv  Detail & Related papers  (2020-05-11T16:29:12Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.