Zilin Wang

The Hong Kong University of Science and Technology (Guangzhou) · China

Unclaimed · AI-structured profileOpenAlex

Early-career HKUST(GZ) researcher spanning LLM-agent skill repair and YOLO-based medical image detection

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-15) 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 scored: OpenAlex reports h-index 0 and i10-index 0, and this record is too new for reliable productivity aggregation. This axis would measure total publication volume and yearly output once the indexed record accumulates more works.

Impact—

Not scored: citation-based influence cannot be estimated — h-index and i10-index are both 0 in the current OpenAlex record, and the indexed works are 2026 preprints too recent to have citations. This axis would measure citation reach and downstream uptake.

Focus0/100

Scores 0, reflecting an unusually even split across two distant areas: the field distribution is exactly 50% natural language processing (LLM agents) and 50% computer vision / medical image analysis, and all five indexed topic labels — advanced neural network applications, sperm and testicular function, NLP techniques, software testing and debugging, and AI-based problem solving and planning — carry a count of 1. With only 2 recent works, concentration metrics have little to aggregate.

Collaboration50/100

Scores 50: the OpenAlex record lists 11 distinct coauthors across recent works, and both catalog papers are middle-author, non-corresponding contributions. The frequent-coauthor set — including Mengyi Deng, Xin Li, Duyi Pan, Zhijiang Guo and Shengqi Chen — spans both the agent-side and vision-side projects, indicating team-embedded work rather than an independent, solo-led research line so far.

Activity32/100

Scores 32, matching a very early record: recent3yWorks is 2, countsByYear contains only 2026 with 2 works, and both catalog papers are 2026 arXiv preprints. Output is entirely concentrated in the current year with no earlier indexed history — real but thin activity, a starting cadence rather than a sustained publication rhythm.

Reproducibility78/100

The strongest axis at 78, computed over 2 reviewed papers (reproMean 78, reproPapers 2). SpermYOLO is rated high — thorough ablations support its reported SVIA AP figures — while RESKILL is medium, with consistent gains across six settings but no statistical testing. Both are arXiv releases, which keeps reported numbers checkable even where the evidence base is still thin.

Research summary

Zilin Wang is an early-career researcher at The Hong Kong University of Science and Technology (Guangzhou) whose indexed record currently consists of two 2026 arXiv preprints in two distinct areas of applied AI: natural language processing (LLM agents) and computer vision / medical image analysis, each exactly 50% of the recorded field distribution. The first, RESKILL, introduces explicit failure attribution and structured, hypothesis-weighted skill repair for interactive language agents; Evolve's review finds the gains consistent across six settings but small and untested statistically. The second, SpermYOLO, is a YOLOv11n-based detector for sperm and impurity detection in microscopic images, lifting SVIA sperm/impurity AP to 97.2/75.4 with thorough ablations and a high reproducibility rating. Both are middle-author, non-corresponding contributions, and the record lists 11 distinct coauthors. OpenAlex reports h-index 0 and i10-index 0 — expected for a record whose only indexed year is 2026 — so output and impact are left unscored rather than estimated. What the data does show clearly is a verification-first orientation: the mean reproducibility score across the 2 reviewed papers is 78, and indexed topics span neural network applications, NLP techniques, software testing and debugging, AI-based planning, and sperm and testicular function. Because 'Zilin Wang' is a common name, this profile is name-resolved from public bibliographic data.

Collaboration pattern

The OpenAlex record lists 11 distinct coauthors across recent works, and both catalog papers are middle-author, non-corresponding contributions — a team-player pattern rather than independent leadership. The frequent-coauthor set spans both research lines: Mengyi Deng, Xin Li, Duyi Pan, Zhiwei Li and Zhijiang Guo appear alongside Wei Wang, Shengqi Chen, Xingyu Pan, Wenting Yu, Pengchao Deng and Guohua Wu, with the collaboration dimension scoring 50. No solo-authored indexed work appears in the current record.

Research keywords

LLM agentsskill repairfailure attributioncredit assignmentself-improvementsperm detectioncomputer-assisted semen analysis (CASA)YOLOv11object detectionmicroscopic image analysisimpurity discrimination

Highlights

Structured skill repair for interactive LLM agents
RESKILL pairs explicit failure attribution with hypothesis-weighted, structured skill repair so language agents can fix their own procedures; Evolve's review notes gains are consistent across six settings (ALFWorld among them) but small and lacking statistical tests. Reproducibility rated medium.
SpermYOLO lifts microscopic sperm detection to 97.2 AP
A YOLOv11n-derived detector coordinates sperm and impurity discrimination in microscopic images, raising SVIA sperm/impurity AP to 97.2/75.4 with thorough ablations; Evolve rates reproducibility high, while noting single-run margins are small and novelty is combinational.
A verification-first, two-field starting record
Across the two 2026 arXiv releases, the mean reproducibility score is 78 (2 papers reviewed), with one high and one medium rating — evidence of habits built around ablations, attribution, and checkable numbers from the outset of an independent record.

Timeline

2026RESKILL preprint released on arXiv: explicit failure attribution and structured, hypothesis-weighted skill repair for interactive language agents (middle-author, non-corresponding).
2026SpermYOLO preprint released on arXiv: a YOLOv11n-based detector for sperm and impurity detection in microscopic images, with SVIA sperm/impurity AP of 97.2/75.4.
2026OpenAlex indexed record begins for this profile: 2 works in 2026, 11 distinct coauthors, affiliation recorded at The Hong Kong University of Science and Technology (Guangzhou).

Papers in the Evolve catalog

Frequent co-authors

Mengyi Deng×1 Xin Li×1 Duyi Pan×1 Zhiwei Li×1 Zhijiang Guo×1 Wei Wang×1 Shengqi Chen×1 Xingyu Pan×1 Wenting Yu×1 Pengchao Deng×1 Guohua Wu×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.

Bibliometric figures (h-index, i10-index, counts by year, coauthor counts) come from OpenAlex only; the record currently reports h-index 0 and i10-index 0.
'Zilin Wang' is a common name and no ORCID is provided; this profile is name-resolved from public bibliographic data and may merge or split identities.
The Evolve catalog holds only two 2026 arXiv preprints; no peer-reviewed venue, funding, or award information is in scope.
Output and impact dimensions are unscored because citation data is absent, not because impact is zero.