Research map / Vision-language and multimodal / Cross-modal learning
BioX-Bridge: Model Bridging for Unsupervised Cross-Modal Knowledge Transfer across Biosignals
Chenqi Li, Yu Liu, Timothy Denison, Tingting Zhu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| select bridge position with the two-stage strategy instead of fixed positions (ablation) | BioX-Bridge · WESAD PPG->ECG | balanced accuracy: 48.34 -> 52.02 | Improved |
| use BioX-Bridge instead of KD with HuBERT-ECG small | unsupervised cross-modal transfer · WESAD PPG->ECG | balanced accuracy: 47.03 -> 52.02, with 0.4M vs 30.4M trainable params | Improved |
| use BioX-Bridge instead of KD-Contrast | cross-modal transfer · ISRUC ECG->EEG | balanced accuracy: 65.92 -> 62.55 | Hurt |
| use BioX-Bridge instead of KD-Contrast | cross-modal transfer · FOG EMG->EEG | balanced accuracy: 68.51 -> 68.04 | No clear effect |
| use BioX-Bridge instead of KD | cross-modal transfer · ISRUC EEG->ECG | balanced accuracy: 60.24 -> 60.11 | No clear effect |
| replace HuBERT-ECG with ECG-FM backbone (ablation) | BioX-Bridge vs KD · WESAD PPG->ECG | balanced accuracy: 48.44 (KD) -> 58.80 | Improved |
| replace HuBERT-ECG with ECG-DualNet (ablation) | BioX-Bridge vs KD · WESAD PPG->ECG | balanced accuracy: 52.78 (KD) -> 56.10 | Improved |
| scale HuBERT-ECG from small to base and large (ablation) | BioX-Bridge and KD · WESAD PPG->ECG | balanced accuracy | Improved |
| scale HuBERT-ECG from small to base and large (ablation) | KD-Contrast · WESAD PPG->ECG | balanced accuracy | Hurt |
| shrink the paired training set to 20% (ablation) | BioX-Bridge · WESAD PPG->ECG | balanced accuracy: about 2% decay at 20% dataset size | Hurt |
| too small or too large bridge rank and prototype set size (ablation) | BioX-Bridge · WESAD PPG->ECG | balanced accuracy | Hurt |
Closest papers by meaning
- Large-scale Training of Foundation Models for Wearable Biosignals (ICLR 2024 · similarity 0.85)
- BioBridge: Bridging Biomedical Foundation Models via Knowledge Graphs (ICLR 2024 · similarity 0.85)
- Scaling Wearable Foundation Models (ICLR 2025 · similarity 0.84)
- BioXArena: Benchmarking LLM Agents on Multi-Modal Biomedical Machine Learning Tasks (NeurIPS 2026 · similarity 0.83)
- Causal Representation Learning from Multimodal Biomedical Observations (ICLR 2025 · similarity 0.83)
- Multimodal Lego: Model Merging and Fine-Tuning Across Topologies and Modalities in Biomedicine (ICLR 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
