Clinical AI evaluation meets solar scientific ML: physics-informed spectral inversion and LLM-vs-physician benchmarking
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
2
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-19) 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: OpenAlex shows only 12 indexed works since 2016 with an h-index of 2, too thin for a reliable measure. This axis would track total, per-year, and first/corresponding-author production as coverage improves.
Impact—
Not yet scored: with an i10-index of 1 and no reliable citation aggregation, impact cannot be estimated. This axis would measure citations, citing works, and uptake of the 2026 arXiv papers once citations accumulate.
Focus0/100
Focus scores 0 because the record is evenly split: field distribution is exactly 0.5 medical informatics / clinical AI evaluation and 0.5 solar physics / scientific machine learning, and OpenAlex topics scatter across face and expression recognition, emotion recognition, image enhancement, and astronomy. The two reviewed papers sit in different fields, so no dominant single direction has consolidated yet — a bimodal, early-stage research identity rather than concentrated specialization.
Collaboration58/100
Collaboration scores 58 from 17 distinct coauthors over the recent three-year window, with 12 recurring collaborators including Jiacheng Xie, Xiaoting Tang, and Yang Yu. Both catalog papers are multi-authored, and positions differ — first author (not corresponding) on the solar PINN paper, middle author on the TCM benchmark — suggesting active but not yet leading-team participation across two separate collaboration networks.
Activity70/100
Activity scores 70 on 10 works in the recent three years. Year-by-year counts show acceleration: 1 work in 2016, 1 in 2021, then 4 in 2024 and 5 in 2025 — the most productive stretch — before 1 indexed work so far in 2026. Two 2026 arXiv papers appear in the Evolve catalog, so the low indexed 2026 count likely reflects OpenAlex lag rather than inactivity.
Reproducibility55/100
Reproducibility scores 55, a mean across the 2 reviewed papers, both tagged medium readiness. The PINN paper ships an analytic forward model (MLSI) inside the network, which aids reuse, but validation stops at MLSI itself; the LLM-TCM benchmark relies on single-run inference and a self-evaluated model, and raises blinding-leakage concerns that limit independent replication.
Research summary
Ziyang Zhang (South China University of Technology, CN; ORCID 0000-0003-3266-7013) presents an early-stage, deliberately two-sided record: the field split is exactly 0.5 medical informatics / clinical AI evaluation and 0.5 solar physics / scientific machine learning. Evolve has reviewed two 2026 arXiv outputs. As first author, Zhang embeds the analytic multilayer spectral inversion (MLSI) forward model in a physics-informed neural network, accelerating Hα 6562.8 Å and Ca II 8542.1 Å inversion by 12–60× — solid engineering, though validated only against MLSI itself. As a middle author, Zhang contributes to a real-world benchmark in which frontier LLMs outscore traditional Chinese medicine physicians under blinded expert review, with the verdict flagging blinding-leakage risk, single-run inference, and a self-evaluated model. Bibliometric coverage is thin: OpenAlex indexes 12 works since 2016 (h-index 2, i10-index 1), 10 in the last three years, with 17 distinct coauthors and 12 recurring collaborators. Both reviewed papers rate medium reproducibility readiness (mean 55). OpenAlex topics also include face and emotion recognition and image enhancement, absent from the reviewed papers, so the profile is name-resolved from public bibliographic data; output and impact axes remain unscored for lack of citation data.
Collaboration pattern
The record lists 17 distinct coauthors in the recent three-year window and 12 recurring collaborators, including Jiacheng Xie, Xiaoting Tang, Yang Yu, Jinpu Li, and Shouli Li. Collaboration (58) reflects breadth across two separate networks: a first-authored solar physics paper and a middle-author clinical AI benchmark, with no corresponding-author role in the reviewed set — an active team participant rather than a group lead.
Research keywords
Highlights
LLMs Outscore TCM Physicians in Blinded Real-Case Benchmark
PINN Embeds MLSI Forward Model for 12–60× Faster Spectral Inversion
A Two-Track 2026: Clinical AI Evaluation and Solar Scientific ML
Both reviewed outputs appear in the same year in distinct fields — a first-authored solar physics method paper and a middle-author clinical AI benchmark — matching the exact 0.5/0.5 field split in the underlying bibliometric data.
Timeline
- 2026First-author arXiv paper embeds an analytic multilayer spectral inversion model in a physics-informed neural network, accelerating Hα and Ca II inversion 12–60×.
- 2026Co-authors a real-world benchmark in which frontier LLMs outscore traditional Chinese medicine physicians under blinded expert review.
- 2026Both reviewed outputs land in the same year across clinical AI evaluation and solar scientific machine learning, defining the profile's two-track shape.
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 derive from OpenAlex only; h-index 2 and i10-index 1 reflect indexed works and may lag arXiv-first output.
- Output and impact dimensions are unscored because citation-side data is insufficient, not because they are zero.
- OpenAlex topics include face and emotion recognition and image enhancement, not represented in the two reviewed papers; the profile is name-resolved from public bibliographic data for a relatively common name.
- OpenAlex counts 1 indexed work for 2026 while the Evolve catalog lists two 2026 arXiv papers, indicating indexing lag.