Research questionCan quantitative attribution metrics show where facial-video rPPG models read pulse signals without indicating heart-rate accuracy?Facial-video remote photoplethysmography models estimate pulse from short clips, while attribution maps are often interpreted as evidence of what the model uses. Localization of skin regions may not correspond to accurate heart-rate estimation. Latest papersRecent research connected to this question, newest first.Cross-Dataset Transfer and Reliability of Explainable Artificial Intelligence for RhythmFormer Remote PhotoplethysmographyEvidence covers eight condition-specific RhythmFormer models trained on NCKU-rPPG under varied illumination, speaking, rotation, and cycling conditions, alongside a UBFC-rPPG reproduction. It compares raw attention, rollout, attention flow, and Beyond Intuition using skin coverage and SaCo, with heart-rate error, waveform correlation, and signal-to-noise ratio; conclusions are limited to these datasets, clips, scenarios, and measures.research paper · Sep 3, 2026