Activity counts measure attention, not model quality. Missing counts are unknown. Publication status follows the linked source; a listed venue alone does not confirm peer review.
Protocol: subject-independent 5-fold CV for both datasets.
- Signals
- ECG, HR, HRV, Text, personal attributes
- Form factor
- Other
- Year
- 2026
- Architecture
HRV Transformer (1 layer, 2 heads, hidden 100) + Qwen3.5-9B-generated per-domain HRV descriptions encoded by Qwen3-Embedding-0.6B (4×1024-d → source projectors + orthogonality loss → 100-d), fused by bidirectional cross-attention (4 heads) + MLP
- License
Code: no LICENSE file in repo, so all rights reserved by default. No weights.
Key results
All results- BLP (HRV-conditioned LLM personalization)Table 4 (source, opens in a new tab)
0.261
MCC ↑
Stress / affect · TILES-2018
HRV features + LLM-generated physiological interpretation
- BLP (HRV-conditioned LLM personalization)Table 4 (source, opens in a new tab)
0.614
MCC ↑
Stress / affect · WESAD
chest ECG-derived HRV · HRV features + LLM-generated physiological interpretation
Results as reported by the authors; not independently reproduced. A SOTA star marks the best value among results reported under the same evaluation protocol; ties go to the earlier publication.
- Type
- FM adaptation
- Modalities
- ECGHRHRVTextpersonal attributes
- Form factor
- Other
- Parameters
- Not reported
- Architecture
- HRV Transformer (1 layer, 2 heads, hidden 100) + Qwen3.5-9B-generated per-domain HRV descriptions encoded by Qwen3-Embedding-0.6B (4×1024-d → source projectors + orthogonality loss → 100-d), fused by bidirectional cross-attention (4 heads) + MLP
- Venue
- arXiv 2610.23555v1 (eess.SP, 2026-10-16)
- Year
- 2026
- DOI
- Not reported
- Weights
- Not released
- Licence
- Code: no LICENSE file in repo, so all rights reserved by default. No weights.
- Caveats
- Not reported
- Datasets
- Added
- Oct 19, 2026
- Updated
- Oct 19, 2026
Abstract
Stress detection based on heart rate variability (HRV) has received increasing attention. However, existing personalized approaches integrate personal attributes with HRV through latent representations, limiting the interpretability of their interactions. In this work, we propose an LLM-enhanced personalized stress detection framework that integrates HRV features and personal attributes through LLM-based semantic interpretation. The proposed framework achieves a 29.9% relative MCC improvement over the strongest baseline on TILES-2018 while providing human-readable descriptions of individual physiological responses. The results highlight the potential of LLM-based semantic interpretation for achieving both effective and interpretable personalized stress detection.
Verbatim fromarXiv· arXiv metadata (abstracts are released under CC0)
Key numbers
How it works
Datasets linked in the hub; pretraining vs evaluation roles follow the paper.
- Architecture
- HRV Transformer (1 layer, 2 heads, hidden 100) + Qwen3.5-9B-generated per-domain HRV descriptions encoded by Qwen3-Embedding-0.6B (4×1024-d → source projectors + orthogonality loss → 100-d), fused by bidirectional cross-attention (4 heads) + MLP
- Inputs
- ECG, HR, HRV, Text, personal attributes
- Approach
- stress · HRV · LLM · personalization · Qwen · cross attention
Devices· 2
Full photo credits and licences are on each device page. An icon means we have no openly licensed or manufacturer-cleared photo of that exact device yet.
Leaderboard rows· 2
All leaderboards| Task | Dataset | Metric | Value | Rank |
|---|---|---|---|---|
| Stress detection (acute / prolonged / affect) | TILES-2018 | Accuracy ↑ | 0.631 | #1 / 1 |
| Stress detection (acute / prolonged / affect) | WESAD | Accuracy ↑ | 0.823 | #2 / 2 |
Datasets· 2
Papers· 1
Cite
APA
Ma, W.-H., Wu, Y.-T., Chou, H.-H., & Lee, C.-C. (2026). Beyond Latent Personalization: Integrating HRV-Conditioned Personalization via LLM for Interpretable Stress Detection. arXiv 2610.23555v1. https://arxiv.org/abs/2610.23555
BibTeX
@misc{ma2026beyond,
title = {{Beyond Latent Personalization: Integrating HRV-Conditioned Personalization via LLM for Interpretable Stress Detection}},
author = {Ma, Wei-Hsiang and Wu, Ya-Tse and Chou, Hsing-Hang and Lee, Chi-Chun},
year = {2026},
howpublished = {arXiv 2610.23555v1},
eprint = {2610.23555},
archivePrefix = {arXiv}
}Built from the hub record. Check it against the publisher before submitting.
History
Added to the ChatPPG research hub