Zi Xu

Shanghai University · China

Unclaimed · AI-structured profileOpenAlex

Shanghai University optimization theorist advancing single-loop extragradient methods with near-optimal guarantees for minimax problems

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 yet scored: the current snapshot lacks the complete publication record — venue breakdown, full output counts beyond the 5 works indexed in 2026, and artifact availability — so research productivity cannot yet be measured reliably for this author.

Impact—

Not yet scored: citation-based influence cannot be assessed — OpenAlex currently reports an h-index and i10-index of 0, unsurprising for 2026 arXiv preprints — so this axis, which would track citations and uptake by other researchers, has no meaningful signal yet.

Focus0/100

Focus measures how concentrated a researcher's output is across fields and topics; the snapshot scores it at 0. That reads oddly against the field distribution, which records exactly two fields — minimax optimization theory and nonconvex minimax optimization — at 0.5 each. OpenAlex topic tags are broader, spanning four topics led by sparse and compressive sensing and stochastic gradient optimization (5 each), suggesting tag dispersion rather than genuinely scattered research.

Collaboration36/100

The collaboration score is 28, reflecting a small footprint: OpenAlex records only 3 distinct coauthors. The frequent-coauthor list centers on Minhao Zhang and Minghao Zhang — two near-identical names that may overlap — and also repeats the author's own entry, an apparent deduplication artifact. On both cataloged 2026 papers Zi Xu sits in the last author position without a corresponding-author flag, indicating a consistent senior-team role within one tight cluster.

Activity52/100

Activity scores 52 on a moderate tempo: all 5 OpenAlex-indexed works fall in 2026, which is also the entire recent three-year window (recent3yWorks = 5), and both Evolve-reviewed items are 2026 arXiv preprints. The record shows a concentrated, recent burst rather than a long publication history, so the mid-range score likely reflects a short observable history and limited indexed coverage.

Reproducibility55/100

Reproducibility averages 55 across 2 reviewed papers. Both catalog entries carry medium reproducibility readiness. The strongly convex–strongly concave paper's proof chain was verified, though constants are loose and experiments are small; the nonconvex–concave paper reports thin experiments alongside its complexity guarantees. Theory-heavy output with checked derivations but limited empirical artifacts yields this mid-level score.

Research summary

Zi Xu is an optimization researcher at Shanghai University whose cataloged work centers on the theory of minimax optimization — collapsing multi-loop algorithmic structures into pure single-loop first-order methods with provable, near-optimal convergence rates. The Evolve catalog holds two 2026 arXiv preprints, both with Zi Xu as last author: one establishes a near-optimal √(κxκy) last-iterate rate for strongly convex–strongly concave minimax via a fixed-parameter single-loop extragradient method with Lyapunov analysis; the other claims the first single-loop method to match multi-loop rates in nonconvex–concave minimax, proving optimal optimization stationarity through a matching lower bound in a restricted oracle class plus best-known game stationarity. OpenAlex records five works in 2026 — the entire recent three-year window — an h-index and i10-index of 0, three distinct coauthors, and a field split of 0.5/0.5 between minimax optimization theory and nonconvex minimax optimization. Radar scores place activity at 52, reproducibility at 55 (the mean over two reviewed papers, both rated medium readiness: verified proof chains, loose constants, thin experiments) and collaboration at 28, while output and impact lack sufficient data to score. This is an early-stage, theory-first profile: influence metrics have not yet accumulated and empirical validation trails the strength of the algorithmic guarantees.

Collaboration pattern

Collaboration is narrow: OpenAlex lists 3 distinct coauthors, and the frequent-coauthor record centers on Minhao Zhang and Minghao Zhang — two near-identical names that may denote overlapping individuals — while also repeating the author's own entry, a deduplication artifact. Both cataloged 2026 papers place Zi Xu in the last author position without the corresponding-author flag, consistent with a fixed, senior role inside one small stable team rather than wide external networking (collaboration score: 28).

Research keywords

minimax optimizationstrongly convex–strongly concave problemsextragradient methodssingle-loop algorithmssingle-loop first-order methodslast-iterate convergenceLyapunov analysisnonconvex–concave optimizationoracle complexitylower boundsstochastic gradient optimizationoptimization theory

Highlights

Near-optimal single-loop extragradient for strongly convex–strongly concave minimax
A pure single-loop, fixed-parameter extragradient method achieves a near-optimal √(κxκy) last-iterate rate for strongly convex–strongly concave minimax optimization; Evolve's review verified the proof chain while noting loose constants and small-scale experiments.
First single-loop method claimed to match multi-loop rates in nonconvex–concave minimax
This work claims to be the first single-loop method matching multi-loop nonconvex–concave rates, with optimal optimization stationarity established via a matching lower bound within a restricted oracle class and best-known game stationarity; experiments remain thin.

Timeline

2026OpenAlex indexes 5 works in 2026, which constitutes the entire recent three-year publication window.
2026Posts the near-optimal pure single-loop extragradient method for strongly convex–strongly concave minimax optimization on arXiv, in the last author position.
2026Posts the single-loop method claimed to match multi-loop rates in nonconvex–concave minimax optimization on arXiv, again as last author.

Papers in the Evolve catalog

Frequent co-authors

Zi Xu×2 Minhao Zhang×2 Huiling Zhang×1 Minhao Zhang×1 Minghao Zhang×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 metrics come from an OpenAlex snapshot only (identity: openalex); citation coverage for 2026 arXiv preprints typically lags, and the recorded h-index and i10-index are both 0.
No ORCID is linked; 'Zi Xu' is a fairly common Chinese name, so this profile is name-resolved from public bibliographic data and identity should be confirmed before reuse.
Only 2 of the indexed works have Evolve reviews; quality and reproducibility statements apply only to that small subset.
Topic and field labels are auto-generated OpenAlex classifications and partially conflict with the reviewed papers' minimax optimization focus.