Abeer Alwan

University of California, Los Angeles · United States

Unclaimed · AI-structured profileORCID OpenAlex

UCLA speech and audio processing researcher; recent work on ASR hallucination control and privacy attacks on TTS models

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—

The output axis is not yet scored in this snapshot: calibrated publication-volume data is insufficient here. It would measure total and per-year research output across the career; the raw record does contain year-by-year counts, but no computed output score is available.

Impact—

The impact axis is not yet scored in this snapshot, so no impact estimate is offered. It would measure citation-based influence such as per-work citations and field-weighted impact; raw aggregates (h-index 39, i10-index 147) exist, but no calibrated score is available.

Focus100/100

Focus scores 100. The field distribution for cataloged work is 100% speech processing / automatic speech recognition, and the OpenAlex topic profile is dominated by speech: 323 works in Speech and Audio Processing, 252 in Phonetics and Phonology Research, 247 in Speech Recognition and Synthesis, and 122 in Voice and Speech Disorders, against only 35 in Music and Audio Processing. Recent papers extend this core into foundation-model reliability and speech-AI security rather than departing from it.

Collaboration53/100

Collaboration scores 53. The snapshot lists 13 distinct coauthors, and the frequent-coauthor set is a tight recurring cluster — Kaiyuan Zhang, Kunlin Cai, Zihang Xiang, Jinghuai Zhang, Fnu Suya, Yuan Tian, Natarajan Balaji Shankar, Zilai Wang, Zihan Wang, and Mohan Shi. On the cataloged AURA paper the researcher appears in the last author position, consistent with senior supervision of a small, stable team rather than broad cross-institution networks.

Activity100/100

Activity scores 100. The record shows 56 works in the most recent three-year window, with annual counts of 17 (2024), 22 (2025), and 17 (2026) — a marked re-acceleration after the dip to 4 works in 2020 and 7 in 2019. The highest annual count in the record is 62 (2021), followed by 37 (2002), indicating sustained multi-decade output with a strong current phase.

Reproducibility100/100

Reproducibility scores 100, with a reproducibility mean of 100 computed across 1 assessed paper — a perfect but very small base. Among the two cataloged papers, AURA carries high reproducibility readiness and Hearing the Whispers medium, the latter limited by single-split evaluation without variance reporting. The axis measures whether released code, data, and evaluation protocols allow independent replication; more assessed papers would firm up the score.

Research summary

Abeer Alwan is a researcher at the University of California, Los Angeles whose OpenAlex-resolved record is anchored in speech and audio processing: 323 works tagged Speech and Audio Processing, 252 in Phonetics and Phonology Research, 247 in Speech Recognition and Synthesis, 122 in Voice and Speech Disorders, and 35 in Music and Audio Processing. The record carries an h-index of 39 and an i10-index of 147, with 56 works in the most recent three-year window and annual counts peaking at 62 in 2021. Radar scores place activity, focus, and reproducibility at 100 each, reflecting a field distribution that is 100% speech processing / automatic speech recognition and a perfect reproducibility mean — though that mean rests on a single assessed paper. Recent cataloged work extends this core into speech foundation-model reliability and security: AURA (2026, last author) introduces uncertainty-routed activation editing that cuts Whisper non-speech hallucination from 89.2% to 1.9% with roughly 500x fewer parameters than LoRA, while Hearing the Whispers (2026) reports the first black-box membership inference attacks on fine-tuned TTS models with AUC above 0.8. Collaboration is moderate (score 53, 13 distinct coauthors), clustered around a recurring UCLA-centered group.

Collaboration pattern

Collaboration in this snapshot is concentrated rather than broad: 13 distinct coauthors, with a recurring core of Kaiyuan Zhang, Kunlin Cai, Zihang Xiang, Jinghuai Zhang, Fnu Suya, Yuan Tian, Natarajan Balaji Shankar, Zilai Wang, Zihan Wang, and Mohan Shi. On the cataloged AURA paper the researcher sits in the last author position, consistent with supervising a small stable team; both 2026 catalog papers involve this same UCLA-centered cluster (collaboration score 53).

Research keywords

speech and audio processingautomatic speech recognitionspeech foundation modelsWhisperhallucination mitigationactivation editingparameter-efficient fine-tuningtext-to-speech synthesismembership inference attackvoice cloningprivacy auditingspeaker verification

Highlights

AURA: activation editing cuts Whisper hallucination from 89.2% to 1.9%
Last-author 2026 arXiv paper introducing uncertainty-routed activation editing for acoustic grounding in speech foundation models. It reduces Whisper non-speech hallucination from 89.2% to 1.9% using roughly 500x fewer parameters than LoRA; Evolve rates reproducibility readiness high but notes the method is Whisper-only.
First black-box membership inference attacks on fine-tuned TTS models
2026 arXiv paper reporting the first black-box membership inference attacks on fine-tuned text-to-speech models, achieving AUC above 0.8 with careful ablations and implications for voice-cloning privacy auditing. Evolve flags medium reproducibility readiness because evaluation uses a single split without variance.

Timeline

1996Record shows 13 publications — the start of a sustained climb (13 in 1996, 13 in 1997, 11 in 1998) after 6 in 1995.
2002Highest annual output in the record before 2021: 37 publications.
2021Peak annual count in the record: 62 publications.
2026Two arXiv papers in the Evolve catalog: AURA (last author) and Hearing the Whispers (middle author).

Papers in the Evolve catalog

Frequent co-authors

Kaiyuan Zhang×2 Kunlin Cai×1 Zihang Xiang×1 Jinghuai Zhang×1 Fnu Suya×1 Yuan Tian×1 Natarajan Balaji Shankar×1 Zilai Wang×1 Zihan Wang×1 Mohan Shi×1 Zihan Wang×1 Natarajan Balaji Shankar×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.

All bibliometrics come from a single OpenAlex-resolved record; citation aggregates (h-index 39, i10-index 147) reflect that source only.
countsByYear contains a 1958 entry (1 work) that predates the rest of the record; this may indicate entity ambiguity or merged records, so this profile is name-resolved from public bibliographic data.
The reproducibility score of 100 rests on only 1 assessed paper; the second catalog paper is rated medium readiness.
Output and impact dimension scores are not computed in this snapshot and are left unscored rather than estimated.