Early-career medical AI researcher at Oxford focused on agentic LLM systems and clinical safety
2
papers here
1
reviewed by Evolve
1
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—
The output axis is not yet computed: the OpenAlex record links only 4 works in the past three years and no complete publication list is available, so total productivity across venues cannot be measured reliably.
Impact—
Impact scoring is unavailable: citation coverage for this early-career record is too sparse (h-index 1, i10-index 0), so the axis, which would measure citations, per-paper influence and field-weighted impact, is left unscored.
Focus100/100
Focus scores 100/100, the strongest signal in this profile: all Evolve-catalogued work sits in one labelled field, medical artificial intelligence (agentic systems), and OpenAlex topic counts concentrate in Artificial Intelligence in Healthcare and Education (3 records), with single records in AI cancer detection, AI ethics, biomedical text mining and multimodal machine learning — an exceptionally tight clinical-AI concentration.
Collaboration39/100
Collaboration scores 39, reflecting a small early-career network: the record lists 6 distinct coauthors, with repeat collaborators concentrated among Junde Wu, Fenglin Liu, Minghao Hu, Jiazhen Pan and Minhao Hu. Both catalogued 2026 preprints are team efforts — MedRSI with Zhu in a middle-author position and the persuadability study as first author (non-corresponding) — indicating active but compact co-authorship rather than a broad distributed network.
Activity47/100
Activity scores 47, consistent with a profile in its take-off phase: counts-by-year show 1 indexed work in 2024 and 3 in 2026, with 4 works recorded in the last three years overall. The 2026 cluster — including two arXiv preprints reviewed by Evolve — suggests output is accelerating, though the short publication window and reliance on preprints temper the score.
Reproducibility55/100
Reproducibility scores 55, averaged over just 1 reviewed paper (reproMean 55). The MedRSI preprint carries a medium reproducibility readiness: Evolve credits strong benchmark gains on glaucoma diagnosis and echocardiography, but flags that the clinical cost measure is judged by an LLM and evaluation relies on private data, both of which limit independent verification.
Research summary
Jiayuan Zhu is an early-career researcher at the University of Oxford working at the intersection of medical artificial intelligence and clinical safety. The OpenAlex-linked record (identity resolved from public bibliographic data; no ORCID provided) shows a compact but highly focused profile: 4 works in the last three years, with 1 indexed output in 2024 and 3 in 2026, and topic clustering dominated by Artificial Intelligence in Healthcare and Education (3 records) alongside single records in AI-based cancer detection, ethics and social impacts of AI, biomedical text mining, and multimodal machine learning. The Evolve-reviewed output centers on agentic medical AI: MedRSI (2026, arXiv), a recursive self-improvement framework for medical agents evaluated on glaucoma diagnosis and echocardiography, which Evolve assesses as claiming the first medical RSI framework with strong benchmark gains, while noting that LLM-judged clinical cost and private evaluation data limit independent verification (reproducibility readiness: medium). A second 2026 first-author preprint, 'AI Can Be Easily Persuaded in Clinical Decision Making', extends the safety agenda to the persuadability of clinical AI. The bibliometric footprint is small so far (h-index 1, i10-index 0, 6 distinct coauthors), consistent with an emerging profile whose focus score of 100 reflects exceptional topical concentration, while output and impact axes remain uncomputed due to insufficient linked data.
Collaboration pattern
Co-authorship is compact: 6 distinct coauthors are recorded, with repeat collaborations concentrated among Junde Wu, Fenglin Liu, Minghao Hu, Jiazhen Pan and Minhao Hu. Both 2026 preprints are collaborative — Zhu is a middle author on MedRSI and first author (non-corresponding) on the persuadability study — suggesting a small, stable core team around medical LLM agent research at Oxford. Note: the coauthor list contains near-duplicate names that may reflect name variants.
Research keywords
Highlights
MedRSI: recursive self-improvement for medical agents
AI Can Be Easily Persuaded in Clinical Decision Making
First-author 2026 arXiv preprint examining the persuadability of AI systems in clinical decision-making; it has not yet been reviewed by Evolve (no verdict or reproducibility rating available), but it extends the profile's clinical-safety focus beyond agentic frameworks.
Timeline
- 2024First indexed output appears in the OpenAlex record (1 work in counts-by-year); no Evolve-reviewed paper is recorded for this year.
- 2026Publishes MedRSI on arXiv in a middle-author position: a recursive self-improvement framework for medical agents, Evolve-reviewed with medium reproducibility readiness.
- 2026First-author arXiv preprint 'AI Can Be Easily Persuaded in Clinical Decision Making' extends the clinical-safety agenda; indexed output in 2026 reaches 3 works.
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.
- Bibliometric data comes from OpenAlex only (identity: openalex); no ORCID is linked, so this profile is name-resolved from public bibliographic records.
- Output and impact radar dimensions are unscored due to insufficient linked data; citation figures are limited to h-index 1 and i10-index 0.
- The reproducibility score is averaged over a single reviewed paper (reproPapers: 1); the second catalogued preprint has no Evolve verdict or reproducibility rating yet.
- The frequent-coauthor list contains near-duplicate names (e.g. 'Jiazhen Pan' appears twice, alongside 'Minghao Hu' and 'Minhao Hu'), suggesting possible name-variant merging in the source record.