Jin Li · Southwest Medical University — LLM on-policy distillation methods and medical AI curriculum design
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-13) 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 computable: the input provides no publication counts, venue history, or yearly output totals for this name. This axis would measure total research output and its distribution over time and across publication venues.
Impact—
Missing from the snapshot: no citation counts, h-index, or downstream uptake are supplied, so influence cannot be scored. This axis would measure citations, usage, and the practical adoption of the distillation method and the teaching system.
Focus0/100
The measured focus score is 0, reflecting an even 0.5/0.5 split in the field distribution between machine learning (LLM post-training and distillation) and medical AI education (curriculum design). With one catalogued paper in each field, the portfolio reads as two parallel tracks rather than one specialized line, although both touch multimodal medical imaging; a larger sample would be needed to tell whether this balance is a deliberate two-track strategy or an artifact of only two assessed papers.
Collaboration39/100
The collaboration score is 39, based on 6 distinct coauthors in the snapshot. Recurring collaborators include Qiangqiang He, MingCai Chen, Dongjing Shan, Yamei Luo, and Yong Luo. On both Evolve-reviewed papers Li holds a middle author position without corresponding-author status, pointing to work inside small teams rather than independent leadership; the network is compact, and coauthor affiliations are not recorded, so cross-institution reach cannot be judged.
Activity—
No activity data is available: the input contains no publication dates beyond the two 2026 preprints, with no yearly volume or submission cadence. This axis would measure recent publication rate and research tempo across consecutive years.
Reproducibility38/100
Reproducibility scores 38 as the mean over 2 assessed papers. Evolve rates the on-policy distillation paper at medium readiness — careful analysis, but no significance tests and evaluation confined to Qwen3 math tasks — and the teaching-system paper at low readiness, since the four-tier curriculum is not yet empirically validated. The pattern suggests conceptual work that is documented but not yet packaged for independent replication.
Research summary
Jin Li is an early-stage researcher at Southwest Medical University whose catalogued work spans two parallel tracks: methodology in large language model post-training and distillation, and pedagogy in medical AI education. Both catalogued outputs are 2026 arXiv preprints reviewed by Evolve. The first, 'Calibrating Teacher–Student Discrepancy for On-Policy Distillation,' argues that teacher self-deviation pollutes the on-policy distillation signal and proposes Cal-OPD, which filters that noise through privileged probes; Evolve credits the analysis as careful but flags missing significance tests and an evaluation limited to Qwen3 math tasks. The second, 'From Research Frontier to Laboratory Bench,' translates endometrial-carcinoma AI research into a four-tier undergraduate experimental teaching system for multimodal medical image intelligent diagnosis, built on constructive alignment and CDIO; Evolve judges it pedagogically sound but not yet empirically validated. On measured dimensions, the snapshot shows a compact collaboration network (score 39; 6 distinct coauthors), a reproducibility mean of 38 across 2 assessed papers with per-paper readiness of low and medium, and an even 0.5/0.5 field split that yields a focus score of 0. Output, impact, and activity metrics are not available in the input. On both papers Li sits in a middle author position without corresponding-author status, consistent with a contributing team member rather than a lab lead.
Collaboration pattern
Jin Li's recorded network is compact: 6 distinct coauthors, giving a collaboration score of 39. The recurring set comprises Qiangqiang He, MingCai Chen, Dongjing Shan, Yamei Luo, and Yong Luo. On both catalogued 2026 papers Li appears in a middle author position and is not the corresponding author, consistent with contributing within a small team rather than leading an independent group; coauthor affiliations are absent from the snapshot, so cross-institution reach cannot be assessed.
Research keywords
Highlights
Cal-OPD: calibrating teacher–student discrepancy in on-policy distillation
Four-tier lab system bringing medical AI research into undergraduate teaching
Timeline
- 2026Co-authored 'Calibrating Teacher–Student Discrepancy for On-Policy Distillation' on arXiv, in a middle author position, proposing the Cal-OPD method.
- 2026Co-authored the design of a four-tier experimental teaching system for multimodal medical image intelligent diagnosis, again in a middle author position.
- 2026Both 2026 outputs entered the Evolve catalog with medium and low reproducibility readiness, defining the current two-track profile.
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 dimension scores are computed from a small bibliometric snapshot only; no citation counts, publication totals, or h-index were available.
- 'Jin Li' is a common name and no ORCID is linked; this profile is name-resolved from public bibliographic data, so identity disambiguation carries some uncertainty.
- Both catalogued papers are 2026 arXiv preprints reviewed by Evolve; verdicts reflect review-stage assessment, not a peer-reviewed publication record.
- Coauthor affiliations and country data are absent from the input, so cross-institution and international collaboration cannot be assessed.