Speech AI researcher building full-duplex dialogue agents and benchmarks for vocal emotion and structured-token ASR.
3
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
Evolve does not yet have sufficient publication-volume data for this axis. It would measure overall research output, such as total indexed works, the output trajectory across years, and productivity relative to field norms.
Impact—
Impact cannot yet be scored from available data. This axis would measure citation-based influence, such as accumulated citation counts, h-index momentum, and uptake of the speech benchmarks by other research groups.
Focus100/100
The focus score is 100, the maximum. Every indexed signal sits in one narrow band: OpenAlex topics are led by Speech Recognition and Synthesis (3) with Speech and dialogue systems (2), plus single counts for Emotion and Mood Recognition, Speech and Audio Processing, and Music and Audio Processing; the only classified field is spoken dialogue systems / speech NLP (1). All three 2026 papers are speech-AI evaluation studies.
Collaboration48/100
The collaboration score is 48, a mid-range value built on 10 distinct coauthors in the OpenAlex record. Yi Zhong appears in the last author position on all three cataloged 2026 papers, a senior-author slot, though none is marked corresponding. Recurring collaborators include Tyler Baumgartner, Brandon Tai, Candice Fan, Luc Debaupte, and Bill Wang, suggesting a small stable team rather than a wide network.
Activity40/100
The activity score is 40. The entire visible record is compressed into one year: countsByYear shows 3 works in 2026 and nothing before it, and recent3yWorks is 3. This reads as an active burst concentrated in 2026 rather than a sustained multi-year cadence, so the score should be revisited once more years of indexed output accrue.
Reproducibility20/100
The reproducibility score is 20, computed from 1 paper carrying reproducibility data with a reproducibility mean of 20. That single assessed paper, the full-duplex adaptation study, is rated low reproducibility readiness, and Evolve's review flags partly circular labels — a concrete obstacle for others attempting to reuse the experimental setup.
Research summary
Yi Zhong is an early-career researcher in speech and audio AI whose OpenAlex-linked record centers on spoken dialogue systems and speech recognition. All three works in Evolve's catalog appeared in 2026, matching the platform's activity signal of 3 recent works over three years and 3 works indexed in 2026. The research program has a strong evaluation orientation: VocalAffectBench targets vocal emotion recognition in AI audio models, and VoiceCodeBench tests exact structured-token recovery in automatic speech recognition. The flagship paper, 'Continue, Adapt, or Yield: In-Turn Adaptation to Overlapping Speech in Full-Duplex Agents' (2026), where Yi Zhong is last author, introduces Duplex Cue, a 3x3 intent-by-response matrix for full-duplex agents; Evolve's review notes humans adapt on 68.2% of collaborative cues versus 34.8% for PersonaPlex, while flagging that labels are partly circular. Bibliometrically the profile is new and tightly concentrated: focus 100, collaboration 48 with 10 distinct coauthors, activity 40, and reproducibility 20, with the one assessed paper rated low reproducibility readiness. OpenAlex reports an h-index and i10-index of 0, and output and impact are treated as lacking sufficient data.
Collaboration pattern
Yi Zhong works with a compact team: 10 distinct coauthors appear in the OpenAlex record, and the frequent-coauthor list repeats Bill Wang alongside Tyler Baumgartner, Brandon Tai, Candice Fan, Luc Debaupte, Yunqi Lu, Nikhil Johri, Ruben Aguilar, and Lisa Kaelin-Martin. Consistent last-author placement across all three 2026 papers suggests a senior or supervisory role, though no paper is flagged as corresponding author.
Research keywords
Highlights
Duplex Cue: a 3x3 intent-by-response matrix for in-turn adaptation in full-duplex agents
VocalAffectBench: evaluating vocal emotion recognition in AI audio models
A 2026 arXiv paper with Yi Zhong in the last author position. It addresses vocal emotion recognition in AI audio models. The catalog record currently carries only the title, year, and source — no Evolve verdict, keywords, or reproducibility rating have been published yet.
VoiceCodeBench: testing exact structured-token recovery in ASR
A 2026 arXiv paper with Yi Zhong in the last author position. It evaluates exact structured-token recovery in automatic speech recognition. As with the other 2026 entry, the catalog record holds only the title, year, and source; no verdict or reproducibility assessment is available yet.
Timeline
- 2026Released 'Continue, Adapt, or Yield: In-Turn Adaptation to Overlapping Speech in Full-Duplex Agents' on arXiv, introducing Duplex Cue; Yi Zhong is last author.
- 2026Released 'VocalAffectBench: Evaluating Vocal Emotion Recognition in AI Audio Models' on arXiv, last author.
- 2026Released 'VoiceCodeBench: Evaluating Exact Structured-Token Recovery in Automatic Speech Recognition' on arXiv, last author.
- 2026All 3 cataloged works land in this single year; OpenAlex countsByYear shows 3 works in 2026 and no earlier indexed output.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- All bibliometrics come from an OpenAlex-linked record; the profile is name-resolved from public bibliographic data, and no institution, country, or ORCID is available to confirm identity.
- Output and impact are unscored because the underlying data is insufficient; OpenAlex currently reports h-index 0 and i10-index 0, which more likely reflects a newly split or sparse author record than a measured impact level.
- Only 1 of the 3 papers has an Evolve verdict, reproducibility rating, and keywords; the VocalAffectBench and VoiceCodeBench records carry titles and years only.
- OpenAlex activity data shows 3 works in 2026 and nothing earlier, so no multi-year productivity trend can be assessed.