Brandon Tai

—

Unclaimed · AI-structured profileOpenAlex

Speech AI researcher: full-duplex dialogue turn-taking, vocal emotion recognition, and structured-token ASR benchmarks

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

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—

No output score is available: the citation profile lists h-index 0 and i10-index 0 with only three records in 2026, so accumulated publication volume cannot be measured yet. This axis would count total and per-year outputs across venues and document types.

Impact—

No impact score is available: with h-index 0, i10-index 0 and works only from 2026, there are no meaningful citation counts to aggregate. This axis would measure citations, citing works, and downstream scholarly influence once citations accrue.

Focus100/100

The focus score is 100/100, the strongest signal in this profile. OpenAlex topic assignments all sit inside speech: Speech Recognition and Synthesis (3 records), Speech and dialogue systems (2), plus one each for Emotion and Mood Recognition, Speech and Audio Processing, and Music and Audio Processing. The field distribution lists a single field, spoken dialogue systems / speech NLP, and all three 2026 catalog papers address speech.

Collaboration48/100

The collaboration score is 48/100. OpenAlex records 10 distinct coauthors, and the frequent-coauthor list includes Tyler Baumgartner, Candice Fan, Luc Debaupte, Yi Zhong, Bill Wang, Yunqi Lu, Nikhil Johri, Ruben Aguilar, and Lisa Kaelin-Martin. On all three catalog papers Brandon Tai is a middle-position author and corresponding author on none, consistent with contributing inside a larger speech-AI team rather than leading as a solo or senior investigator.

Activity40/100

The activity score is 40/100. OpenAlex countsByYear shows 3 works recorded in 2026 with no entries in any earlier year, and recent3yWorks is also 3 — every visible output lands in a single year, matching the three 2026 arXiv preprints in the catalog. This is consistent with a very new entrant to recorded publication, so the score reflects a short observable window rather than a slowdown.

Reproducibility20/100

The reproducibility score is 20/100, computed as a mean of 20 over 1 assessed paper. The one Evolve-reviewed paper, on in-turn adaptation in full-duplex agents, carries a reproReadiness rating of low, and the review notes that part of its behavioral labels are circular. The two benchmark preprints have not yet been assessed, so this axis rests on a single data point.

Research summary

Brandon Tai is an early-career speech AI researcher whose recorded output — three arXiv preprints, all dated 2026 — sits squarely in spoken dialogue systems and speech evaluation. This profile is name-resolved from public OpenAlex bibliographic data; no institutional affiliation, country, or ORCID is available. The Evolve catalog lists three papers, in all of which Brandon Tai is a middle-position author and none as corresponding author: "Continue, Adapt, or Yield: In-Turn Adaptation to Overlapping Speech in Full-Duplex Agents" (2026), plus two evaluation resources — VocalAffectBench on vocal emotion recognition in AI audio models, and VoiceCodeBench on exact structured-token recovery in automatic speech recognition. The one paper Evolve has reviewed introduces Duplex Cue, a 3×3 intent-by-response matrix for in-turn adaptation to overlapping speech; its reported result is that humans adapt on 68.2% of collaborative cues versus 34.8% for PersonaPlex, though the review cautions that labels are partly circular and rates reproducibility readiness low. Radar dimensions reflect the early stage: focus 100/100, with all OpenAlex topics clustering in speech; collaboration 48/100 over 10 distinct coauthors, including Tyler Baumgartner, Candice Fan, Luc Debaupte, Yi Zhong, and Bill Wang; activity 40/100, with 3 works all in 2026; and reproducibility 20/100 from a single assessed paper. Output and impact are unscored — h-index and i10-index are both 0.

Collaboration pattern

OpenAlex records 10 distinct coauthors. Recurring names include Tyler Baumgartner, Candice Fan, Luc Debaupte, Yi Zhong, Bill Wang, Yunqi Lu, Nikhil Johri, Ruben Aguilar, and Lisa Kaelin-Martin. The pattern — three 2026 preprints, each with Brandon Tai in a middle author position and none as corresponding author — suggests contribution within a larger speech-AI team rather than solo or senior-led work. No affiliation data is available, so the team context cannot be confirmed.

Research keywords

full-duplex spoken dialogueturn-takingoverlapping speechin-turn adaptationspeech language modelsspoken dialogue systemsspeech recognition evaluationstructured-token recoveryvocal emotion recognitionemotion and mood recognitionspeech and audio processingspeech benchmarks

Highlights

Duplex Cue: in-turn adaptation to overlapping speech in full-duplex agents
Evolve's review describes a 3×3 intent-by-response cue matrix and reports humans adapting on 68.2% of collaborative cues versus 34.8% for PersonaPlex, while cautioning that labels are partly circular and rating reproducibility readiness low.
VocalAffectBench: probing vocal emotion recognition in AI audio models
This 2026 arXiv preprint targets vocal emotion recognition in AI audio models and extends the profile's speech focus into affective evaluation. Evolve has not yet published a verdict, keywords, or a reproducibility assessment for it.
VoiceCodeBench: exact structured-token recovery in ASR
This 2026 arXiv preprint evaluates whether automatic speech recognition systems recover structured tokens exactly, adding an evaluation-resource strand to the record. No Evolve verdict, keywords, or reproducibility assessment is available yet.

Timeline

2026First visible output year in the record: OpenAlex countsByYear lists 3 works in 2026 with no earlier entries, and recent3yWorks is also 3.
2026Co-authored "Continue, Adapt, or Yield: In-Turn Adaptation to Overlapping Speech in Full-Duplex Agents", introducing the Duplex Cue 3×3 intent-by-response matrix (middle author, non-corresponding).
2026Co-authored VocalAffectBench, a benchmark for vocal emotion recognition in AI audio models.
2026Co-authored VoiceCodeBench, evaluating exact structured-token recovery in automatic speech recognition.

Papers in the Evolve catalog

2026-09-11
Continue, Adapt, or Yield: In-Turn Adaptation to Overlapping Speech in Full-Duplex Agents

Frequent co-authors

Tyler Baumgartner×3 Candice Fan×3 Luc Debaupte×3 Yi Zhong×3 Bill Wang×2 Yunqi Lu×1 Nikhil Johri×1 Ruben Aguilar×1 Bill Wang×1 Lisa Kaelin-Martin×1

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.

Bibliometrics come from OpenAlex only; identity is name-resolved from public bibliographic data, with no institutional affiliation, country, or ORCID available.
Output and impact are unscored: h-index and i10-index are both 0, so no citation or cumulative-output picture exists yet.
Only one of the three catalog papers has an Evolve review; VocalAffectBench and VoiceCodeBench lack verdicts, keywords, and reproducibility assessments.
All catalog entries are 2026 arXiv preprints; the data contains no peer-reviewed venue information.