Piao, Ruoxi, Lu, Alicia, Zhang, Yunzhen, Mısır, Mustafa, Keleş, Hasan Onur · Oct 28, 2026
BridgeRouter routes biosignal instances among cross-modal pathways
Markdown · JSON · Original source
In their paper, Ruoxi Piao and coauthors argue that choosing one cross-modal bridge for every case can miss routes that work better for particular instances. They introduce BridgeRouter, which predicts whether an alternative pathway offers a gain over a protected default, the Single Best Solver (SBS). At inference, the router uses only the target modality and executes one pathway; it does not need EEG, task labels, or predictions from every candidate route. The authors evaluate seven EEG bridge pathways on EMG-based freezing-of-gait detection and ECG-based sleep classification. Across five seeds, they report balanced accuracy of 81.38% on FOG and 62.92% on ISRUC, improvements of 11.49 and 3.83 percentage points over the respective SBS routes. These results suggest a way to use complementary errors among pathways without running all of them for each instance, though the abstract reports only these two evaluation settings. The gap between the reported results and per-instance bridge oracles also indicates remaining headroom within the evaluated portfolios. Whether this routing approach works across other tasks and biosignal modalities remains an open question.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
cs.AI · Article
Source details
Original source date: Oct 28, 2026 (2026-10-28).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-10T19:35:11.635Z. First observed: 2026-10-30T06:45:28.576Z. Retrieved: 2026-10-30T18:03:27.963Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
The abstract describes evaluation on two settings: EMG-based freezing-of-gait detection on FOG and ECG-based sleep classification on ISRUC; it does not establish performance on other tasks or modalities. The supplied evidence is the abstract and does not provide further experimental details or evidence of real-world clinical outcomes.
Open question
Does BridgeRouter retain its reported advantage on tasks and biosignal modalities beyond the two evaluated settings?
Source-stated
The authors propose BridgeRouter to select among cross-modal pathways at the instance level, predicting route-specific gains relative to a Single Best Solver default.
formulate multi-bridge transfer as an instance-level algorithm-selection problem and introduce BridgeRouter, a gain-aware router over a portfolio of frozen cross-modal pathways.
Source-stated
At inference, BridgeRouter uses the target modality alone and executes one pathway, selecting an alternative only when its predicted gain exceeds a calibrated threshold.
At inference, BridgeRouter observes only the target modality, requires neither EEG, task labels, nor predictions from all candidate routes, and executes exactly one bridge pathway
Source-stated
Across five seeds, the paper reports balanced accuracy of 81.38% on FOG and 62.92% on ISRUC, improvements over the respective SBS routes.
Across five seeds, BridgeRouter achieves $81.38\pm1.11\%$ balanced accuracy on FOG, improving over the BIOT SBS by $11.49$ percentage points, and $62.92\pm2.17\%$ on ISRUC, improving over the CBraMod SBS by $3.83$ points.
Source-stated
The authors report a gap between BridgeRouter's results and per-instance bridge-oracle performance in both evaluated settings.
Per-instance bridge oracles reach $99.96\%$ and $78.61\%$, respectively, revealing substantial remaining portfolio headroom.
Source and contributor context
Piao, Ruoxi, Lu, Alicia, Zhang, Yunzhen, Mısır, Mustafa, Keleş, Hasan Onur · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Foundation model
A large model trained on broad data that can be adapted to many different tasks.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Editorial provenance
Version 1. Publication ID: cf1c793a-a545-44f3-817a-f3a9e821bbed.
Prepared by server-excerpt-repair/stream-excerpt-repair-v1 (original draft: gpt-6-luna); grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-10T19:35:11.635Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: ade594c658ea8c67c8ec8af153d5202d71b5a789bd5eca0effded22b7bbacf3f. Source text hash: 5dff165d64de1b7562b56c934ffc6b6659f081492d63571c4a0dfc3c5b2cdbe7.
Revision and source-check history
Version 1 · 2026-11-10T19:35:11.635Z
BridgeRouter routes biosignal instances among cross-modal pathways
In their paper, Ruoxi Piao and coauthors argue that choosing one cross-modal bridge for every case can miss routes that work better for particular instances. They introduce BridgeRouter, which predicts whether an alternative pathway offers a gain over a protected default, the Single Best Solver (SBS). At inference, the router uses only the target modality and executes one pathway; it does not need EEG, task labels, or predictions from every candidate route. The authors evaluate seven EEG bridge pathways on EMG-based freezing-of-gait detection and ECG-based sleep classification. Across five seeds, they report balanced accuracy of 81.38% on FOG and 62.92% on ISRUC, improvements of 11.49 and 3.83 percentage points over the respective SBS routes. These results suggest a way to use complementary errors among pathways without running all of them for each instance, though the abstract reports only these two evaluation settings. The gap between the reported results and per-instance bridge oracles also indicates remaining headroom within the evaluated portfolios. Whether this routing approach works across other tasks and biosignal modalities remains an open question.
Original source
cs.AI · Article
Source details
Original source date: Oct 28, 2026 (2026-10-28).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-10T19:35:11.635Z. First observed: 2026-10-30T06:45:28.576Z. Retrieved: 2026-10-30T18:03:27.963Z.
Limits
The abstract describes evaluation on two settings: EMG-based freezing-of-gait detection on FOG and ECG-based sleep classification on ISRUC; it does not establish performance on other tasks or modalities. The supplied evidence is the abstract and does not provide further experimental details or evidence of real-world clinical outcomes.
Open question
Does BridgeRouter retain its reported advantage on tasks and biosignal modalities beyond the two evaluated settings?
Source-stated
The authors propose BridgeRouter to select among cross-modal pathways at the instance level, predicting route-specific gains relative to a Single Best Solver default.
formulate multi-bridge transfer as an instance-level algorithm-selection problem and introduce BridgeRouter, a gain-aware router over a portfolio of frozen cross-modal pathways.
Source-stated
At inference, BridgeRouter uses the target modality alone and executes one pathway, selecting an alternative only when its predicted gain exceeds a calibrated threshold.
At inference, BridgeRouter observes only the target modality, requires neither EEG, task labels, nor predictions from all candidate routes, and executes exactly one bridge pathway
Source-stated
Across five seeds, the paper reports balanced accuracy of 81.38% on FOG and 62.92% on ISRUC, improvements over the respective SBS routes.
Across five seeds, BridgeRouter achieves $81.38\pm1.11\%$ balanced accuracy on FOG, improving over the BIOT SBS by $11.49$ percentage points, and $62.92\pm2.17\%$ on ISRUC, improving over the CBraMod SBS by $3.83$ points.
Source-stated
The authors report a gap between BridgeRouter's results and per-instance bridge-oracle performance in both evaluated settings.
Per-instance bridge oracles reach $99.96\%$ and $78.61\%$, respectively, revealing substantial remaining portfolio headroom.
Source and contributor context
Piao, Ruoxi, Lu, Alicia, Zhang, Yunzhen, Mısır, Mustafa, Keleş, Hasan Onur · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Foundation model
A large model trained on broad data that can be adapted to many different tasks.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Editorial provenance
Version 1. Publication ID: cf1c793a-a545-44f3-817a-f3a9e821bbed.
Prepared by server-excerpt-repair/stream-excerpt-repair-v1 (original draft: gpt-6-luna); grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-10T19:35:11.635Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: ade594c658ea8c67c8ec8af153d5202d71b5a789bd5eca0effded22b7bbacf3f. Source text hash: 5dff165d64de1b7562b56c934ffc6b6659f081492d63571c4a0dfc3c5b2cdbe7.
uncertain · 2026-11-10T19:35:11.635Z · anthropic/claude-opus-5.5
Publication: cf1c793a-a545-44f3-817a-f3a9e821bbed
attribution — pass: The draft attributes the paper to Ruoxi Piao and coauthors on arXiv. It keeps the results as the authors' reported findings rather than presenting them as established fact.
Authors:Ruoxi Piao, Alicia Lu, Yunzhen Zhang, Mustafa Mısır, Hasan Onur Keleş
summary — uncertain: The summary is faithful but hard for a non-specialist to follow. It omits the core premise: lightweight bridges let EEG foundation models, which are pretrained on brain signals, run on other biosignals such as muscle (EMG) and heart (ECG) signals without retraining. It also leaves central terms unexplained: cross-modal bridge, target modality, foundation model, balanced accuracy, seeds, per-instance oracle, and the dataset names FOG and ISRUC. Smallest repair: add one plain sentence stating that premise and briefly gloss these terms or link them to term-library tooltips.
Lightweight representation bridges make pretrained, task-adapted EEG foundation models executable on other biosignal modalities without retraining their backbones.
date — pass: The originalPublished value of 2026-10-28 matches the v1 submission date on arXiv. The feed date of 2026-10-30 is later and correctly is not used as the publication date.
[Submitted on 28 Oct 2026]
support — pass: The figures, the improvements over each SBS, the single-route inference, the threshold rule and the oracle headroom all match the abstract. The inference about exploiting complementary errors without running every route mirrors the authors' own conclusion. The limitations and the open question stay appropriately scoped to the two evaluated settings.
anchor-relative gain prediction can convert complementary cross-modal bridge behavior into deployable instance-level routing improvements while retaining single-route inference
media — pass: The selected media is the verified direct link to the original arXiv abstract page, which is the most relevant source access.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivA large model trained on broad data that can be adapted to many different tasks.
Definition referencearXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivA large model trained on broad data that can be adapted to many different tasks.
Definition referencearXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXiv