Search:
Match:
2 results
Research Paper#Medical AI, ECG Analysis, Deep Learning🔬 ResearchAnalyzed: Jan 3, 2026 16:07

ECG Generalization with Morphology-Rhythm Disentanglement

Published:Dec 29, 2025 10:14
1 min read
ArXiv

Analysis

This paper addresses the challenge of generalizing ECG classification across different datasets, a crucial problem for clinical deployment. The core idea is to disentangle morphological features and rhythm dynamics, which helps the model to be less sensitive to distribution shifts. The proposed ECG-RAMBA framework, combining MiniRocket, HRV, and a bi-directional Mamba backbone, shows promising results, especially in zero-shot transfer scenarios. The introduction of Power Mean pooling is also a notable contribution.
Reference

“ECG-RAMBA achieves a macro ROC-AUC ≈ 0.85 on the Chapman--Shaoxing dataset and attains PR-AUC = 0.708 for atrial fibrillation detection on the external CPSC-2021 dataset in zero-shot transfer.”

PermalinkArXiv
Research Paper#Vision-Language Models (VLMs)🔬 ResearchAnalyzed: Jan 3, 2026 16:31

Bi-directional Perceptual Shaping for Improved VLM Reasoning

Published:Dec 26, 2025 18:59
1 min read
ArXiv

Analysis

This paper addresses the limitations of current Vision-Language Models (VLMs) in utilizing fine-grained visual information and generalizing across domains. The proposed Bi-directional Perceptual Shaping (BiPS) method aims to improve VLM performance by shaping the model's perception through question-conditioned masked views. This approach is significant because it tackles the issue of VLMs relying on text-only shortcuts and promotes a more robust understanding of visual evidence. The paper's focus on out-of-domain generalization is also crucial for real-world applicability.
Reference

“BiPS boosts Qwen2.5-VL-7B by 8.2% on average and shows strong out-of-domain generalization to unseen datasets and image types.”

PermalinkArXiv