University of Melbourne NLP researcher building mental-health counseling datasets and LLM safety benchmarks
2
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
1
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—
Not yet scored: the OpenAlex record for this researcher is sparse (h-index 1, i10-index 0, only one year of countsByYear data). This axis would measure total publication volume, venue quality and the output trajectory, none of which can be judged from the current snapshot.
Impact—
Beyond an i10-index of 0, no citation data is available, so impact is unscored. This axis would capture citations, adoption of the released datasets and benchmarks, and influence on mental-health NLP evaluation practice — none yet measurable for this early-stage record.
Focus100/100
The strongest axis (100/100): OpenAlex topic counts are tightly clustered — Artificial Intelligence in Healthcare and Education (7), Explainable AI (7), Ethics and Social Impacts of AI (5), Machine Learning in Healthcare (4), Topic Modeling (3) — and the field distribution lists a single field, mental-health conversational AI. Both cataloged 2026 papers are mental-health NLP, indicating a deliberate focus on clinical dialogue and AI safety rather than scattered output.
Collaboration63/100
OpenAlex records 23 distinct coauthors across 14 recent works, with 12 recurring collaborators including Stephanie Fong, Yiwen Jiang and Jiahe Liu — a broad but young network. On both reviewed 2026 papers Dwyer holds middle and last author positions without corresponding status, consistent with a contributing team member in group-led NLP projects; the 63/100 score reflects breadth of ties not yet consolidated into a lead-authorship pattern.
Activity79/100
Activity is the second-strongest signal (79/100): 14 works in the last three years, with OpenAlex countsByYear attributing all 14 to 2026, and two arXiv preprints already in the Evolve catalog that year. This recent, front-loaded cadence — a counseling dataset and a safety benchmark appearing together — fits an intensive early-project burst, though the absence of earlier yearly counts prevents judging whether the pace is sustained.
Reproducibility55/100
Reproducibility is scored 55/100 from only one assessed paper (reproPapers 1, reproMean 55). Both reviewed papers carry a medium readiness rating: StratCBT's 9.7K-session dataset is fully LLM-synthetic with GPT-4o as judge, and SDARE-Bench relies on synthetic self-labeled data with moderate classifier F1. Artifacts are likely reusable, but the realism and label-quality caveats in Evolve's reviews temper confidence, and more assessed papers are needed to stabilize this estimate.
Research summary
Dominic Dwyer is an early-career natural language processing researcher at The University of Melbourne whose work sits at the intersection of mental-health conversational AI and LLM safety evaluation. The OpenAlex snapshot is thin — h-index 1, i10-index 0, and 14 works in the past three years, all attributed to 2026 — so this profile is name-resolved from public bibliographic data rather than a verified biography. Topic counts concentrate on Artificial Intelligence in Healthcare and Education (7), Explainable AI (7), Ethics and Social Impacts of AI (5), Machine Learning in Healthcare (4) and Topic Modeling (3). Two 2026 arXiv preprints anchor the record: StratCBT, the first large strategy-aligned CBT counseling dataset (9.7K sessions), which Evolve flags as valuable but fully LLM-synthetic and GPT-4o-judged; and SDARE-Bench, the first benchmark for LLM stigma detection in dyadic and group dialogue, whose group-pressure results are notable even though synthetic self-labeled data and moderate classifier F1 weaken its conclusions. Both papers rate medium reproducibility readiness (reproMean 55 from one assessed paper). He collaborates widely — 23 distinct coauthors, 12 recurring — while holding middle and last authorship without corresponding status.
Collaboration pattern
Dwyer's OpenAlex record lists 23 distinct coauthors across 14 recent works, with 12 recurring collaborators — including Stephanie Fong, Yiwen Jiang, Jiahe Liu, Zimu Wang and Hongxi Yang — indicating a dense, early-career Melbourne-centric NLP cluster. On both reviewed 2026 papers he is neither first nor corresponding author (middle on StratCBT, last on SDARE-Bench), consistent with a contributing specialist in group-led projects rather than a lab head; expect team-based dataset and benchmark work.
Research keywords
Highlights
StratCBT: first large strategy-aligned CBT counseling dataset
SDARE-Bench: first stigma benchmark for LLM group dialogue
Timeline
- 2026StratCBT preprint appears on arXiv, introducing a 9.7K-session dataset that aligns psychological counseling conversations with CBT strategies (middle authorship).
- 2026SDARE-Bench preprint appears on arXiv, the first LLM benchmark for stigma detection and response in dyadic and group dialogue (last authorship).
- 2026OpenAlex attributes all 14 of the past-three-years works to this single year, with 23 distinct coauthors recorded across the recent output.
- 2026Reproducibility assessment begins in the Evolve catalog: one paper assessed at medium readiness, giving reproMean 55 from a single paper.
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.
- Bibliometrics come from OpenAlex only; the record is sparse (h-index 1, i10-index 0, countsByYear limited to 2026).
- No ORCID is provided; 'Dominic Dwyer' is name-resolved from public bibliographic data, and confusion with other same-named researchers cannot be fully excluded.
- Only two papers, both 2026 arXiv preprints, are reviewed in the Evolve catalog; peer-reviewed venues and earlier career history are not covered.
- The reproducibility score rests on a single assessed paper (reproPapers 1), so the 55 estimate is fragile.