Ming Lei

Renmin University of China · China

Unclaimed · AI-structured profileOpenAlex

XAI researcher at Renmin University of China, bridging post-hoc interpretability methods and human-centered explanation

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-14) 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—

Output is not yet scoreable: the OpenAlex-linked snapshot indexes only two works, both from 2026, which is too small a base for any volume or productivity measure. This axis would track publication count, venue spread, and authorship position as the record grows.

Impact—

Impact cannot be scored yet: h-index and i10-index both read zero in this snapshot and citation aggregates are unavailable. This axis would measure citation uptake, the quality of citing works, and downstream influence once the indexed record matures.

Focus0/100

Focus scores 0/100, driven by an even field distribution — 0.5 machine learning / explainable AI and 0.5 human-centered explainable AI — rather than concentration in one line. Topic frequencies mirror this: explainable AI leads with 2 papers, while software testing and debugging, data visualization and analytics, AI ethics, and ML classification each appear once. The record is coherent around XAI but too small to show a settled specialty.

Collaboration32/100

Collaboration scores 32/100 from a compact network: the snapshot records 4 distinct coauthors, all recurring — Lemen Chao, Anran Fang, Zixuan Yang, and Mingran Sun. Both reviewed papers are team efforts, with Ming Lei in middle authorship on the ADORE methods paper and last authorship on the data-storytelling study, though neither lists a corresponding-author role. The pattern suggests a stable, small intra-institutional group rather than wide external collaboration.

Activity32/100

Activity scores 32/100, consistent with a newly visible record: countsByYear lists exactly 2 works, both in 2026, and recent3yWorks is also 2. Both entries are arXiv preprints from that single year, so current publishing cadence is real but brief — there is no multi-year history to indicate sustained pace. As later years accumulate, this axis should clarify whether the 2026 burst marks the start of steady output.

Reproducibility55/100

Reproducibility scores 55/100, computed as a mean of 55 across 2 assessed papers — both rated medium readiness by Evolve's review. The ADORE interpretability framework has plausible efficiency gains but flawed adaptivity logic and no significance tests, while the storytelling study reports significant survey gains from a single case with small, uncorrected effects. Methodological transparency is partial: promising, but not yet independently verifiable.

Research summary

Ming Lei is an early-career researcher at Renmin University of China whose recorded output centers on explainable artificial intelligence (XAI). This profile is name-resolved from OpenAlex bibliographic data; the snapshot is small, with two indexed works in 2026 (countsByYear shows 2 in 2026, and recent3yWorks is 2), so citation-based signals such as h-index and i10-index currently read zero. The work splits evenly across two subfields — machine learning / explainable AI and human-centered explainable AI (fieldDist 0.5/0.5) — and Evolve has reviewed both papers. The first, ADORE (2026, arXiv), proposes a unified framework for global and local interpretability via adaptive derivative-ordered random explanation, combining finite differences, Hessian/Jacobian approximation, and randomized SVD; Evolve judged its efficiency gains plausible while flagging flawed adaptivity logic and absent significance testing. The second applies data storytelling to interpretable machine learning, converting SHAP outputs into ABT narrative structures for non-experts without exposing sensitive data or model details; the survey evidence showed gains but rests on a single case with small, uncorrected effects. Both papers carry a medium repro-readiness rating (reproMean 55 across reproPapers 2). Topic frequencies reinforce the XAI core (2) with single appearances in software testing, data visualization, AI ethics, and ML classification.

Collaboration pattern

The recorded collaboration footprint is small and stable: 4 distinct coauthors, all recurring — Lemen Chao, Anran Fang, Zixuan Yang, and Mingran Sun. Both 2026 arXiv papers are co-authored group outputs, with Ming Lei placed middle on the ADORE methods paper and last on the data-storytelling study, without a flagged corresponding role in either. The pattern points to a close-knit team at Renmin University of China rather than extended external networks; no cross-institution partners appear in this snapshot.

Research keywords

explainable AIpost-hoc interpretabilityfeature attributionfinite differencesHessian and Jacobian approximationrandomized SVDinterpretable machine learningdata storytellingSHAPABT narrative structurelarge language modelshuman-centered AI explanation

Highlights

ADORE: a unified global–local interpretability framework
The 2026 arXiv paper combines adaptive derivative-ordered random explanation with finite differences, Hessian/Jacobian approximation, and randomized SVD to unify global and local post-hoc interpretability; Evolve finds the efficiency gains credible but flags flawed adaptivity logic and missing significance tests.
Data storytelling that explains AI decisions to non-experts
The second 2026 arXiv paper converts SHAP explanations into ABT narrative structures, using large language models to decode AI decisions for lay audiences without exposing sensitive data or model details; Evolve notes significant survey gains but single-case evidence with small, uncorrected effects.

Timeline

2026Recorded output begins: the OpenAlex-linked countsByYear attributes 2 works to 2026, both arXiv preprints.
2026Posts the ADORE preprint on arXiv, unifying global and local interpretability via adaptive derivative-ordered random explanation with randomized SVD.
2026Posts the data-storytelling study on arXiv, turning SHAP outputs into ABT narratives that explain AI decisions to non-experts.

Papers in the Evolve catalog

Frequent co-authors

Lemen Chao×2 Anran Fang×2 Zixuan Yang×1 Mingran Sun×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 (countsByYear, h-index, i10-index, topics) come from an OpenAlex identity resolution without an ORCID link; Ming Lei is a common name, so records may be incomplete or partly misattributed.
Only 2 works, both 2026 arXiv preprints, underlie this profile; the output and impact dimensions could not be scored.
The reproducibility score reflects Evolve's review of just 2 papers (mean 55), too small a base for a general claim.
The frequent-coauthor list is limited to the 4 names in the snapshot; coauthor affiliations were not provided.