Local or Escalate? Post-Training Expert Routing for Real-Time Duplex Speech Models
Source check: Source checked. Not independent scientific confirmation.
Finding: A probe trained on typed questions detected speech-answer failures in unseen human recordings, an experimentally measured cross-modal result.
Technology: Logistic probe on language-model hidden states for speech-answer failure detection
Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- A logistic readout on typed-question hidden states was applied to speech-question states without refitting.
- Test and result
- Primary results report discrimination on unseen human speech recordings, beyond the system-level routing result.
- Limitation / still unknown
- Evaluation was on selected internal/external speech-QA datasets; it does not establish general calibration.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://arxiv.org/html/2611.01639v1
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.