Paper
OPLR: Alleviating Normality Bias in PPG-to-ABP Reconstruction via Ordered Prototype Latent Refinement
ICLR 2027 submission (under review)
TL;DR
Names and formalizes Normality Bias: long-tailed BP distributions make PPG→ABP models regress toward the normotensive mean (flattened hypertensive peaks, raised hypotensive troughs) even when aggregate MAE looks good. Introduces two model-agnostic diagnostics — Normality Bias Index (NBI) (monotonic regime-dependent drift) and TailBias (error at clinical extremes) — and shows the bias is pervasive across architecture families. OPLR imposes an ordered Low < Normal < High latent geometry with routing-guided prototype–token refinement.
- Form factor: finger
Results on ChatPPG leaderboards
All leaderboardsNo results from this paper on ChatPPG leaderboards yet. Browse the leaderboards.
Tasks & signals
Tasks
Signals
Keywords
Datasets
How it works
Cite this paper
APA
OPLR: Alleviating Normality Bias in PPG-to-ABP Reconstruction via Ordered Prototype Latent Refinement. (2026). ICLR 2027 submission. https://openreview.net/forum?id=5aEQxcBV3m
BibTeX
@misc{oplrnormalitybia2026oplr,
title = {{OPLR: Alleviating Normality Bias in PPG-to-ABP Reconstruction via Ordered Prototype Latent Refinement}},
year = {2026},
howpublished = {ICLR 2027 submission},
url = {https://openreview.net/forum?id=5aEQxcBV3m}
}Built from the hub record. Check it against the publisher before submitting.
Added to the ChatPPG research hub
