Jiheng Zhang

Hong Kong University of Science and Technology · Hong Kong SAR China

Unclaimed · AI-structured profileORCID OpenAlex

Theory-driven operations researcher at HKUST spanning revenue-management online learning and LLM inference serving.

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-17) 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 platform lacks sufficient verified data on total publication output. This axis would measure cumulative indexed works and annual production; it is left blank rather than estimated, partly because homonymous-author records may inflate counts under a common name.

Impact—

Not yet scored: citation evidence is insufficient for a reliable impact measure. This axis would capture citations, h-index trajectory and per-paper influence; the snapshot lists h-index 21 and i10-index 34, but they are not treated as a verified basis.

Focus0/100

Scores 0, the lowest concentration. OpenAlex topics spread across five clusters — advanced queuing theory (27), glioma diagnosis and treatment (19), optimization and search (17), ferroptosis and cancer prognosis (11), probability and risk models (11) — and fieldDist splits evenly, 0.5 machine-learning systems versus 0.5 operations research. Such dispersion partly reflects homonymous records, so the extreme score should be read cautiously.

Collaboration39/100

Scores 39, a modest collaboration footprint. The last three years show 30 works with only 6 distinct coauthors, and the recurring pool is compact — Zean Han, Zezhen Ding, Jing Liang, Ruihan Lin, Zhenghong Huang and Hongfan Wu. The three Evolve-reviewed 2026 preprints place this researcher in the last author position, consistent with a senior supervisory role in a small, stable team rather than a large distributed network.

Activity100/100

Scores 100, the top activity tier. Annual counts in the record jump from 4 works in 2021 to 19 in 2022, then 15 in 2023, 10 in 2024, 14 in 2025 and 6 already in 2026, with 30 works in the recent three-year window. Recent output centers on arXiv preprints in LLM inference serving and revenue management, indicating a sharply accelerated and currently very active research program.

Reproducibility55/100

Scores 55, a middling reproducibility readiness, computed from 2 Evolve-reviewed papers with a mean of 55, each rated medium. The verdicts flag concrete gaps: a reported 2.5% FWP gain over 88 prompts without variance estimates, proofs described as sketches, experiments lacking baselines, and evaluations limited to synthetic data. Theory construction is consistently strong; empirical artifact completeness is the recurring weakness.

Research summary

Jiheng Zhang is a researcher at Hong Kong University of Science and Technology whose cataloged work splits evenly, 0.5 and 0.5 in the field distribution, between operations research and online learning for revenue management and machine-learning systems for LLM inference serving. The OpenAlex record shows deep applied-probability roots: advanced queuing theory is the largest topic cluster (27 works), alongside optimization and search problems (17) and probability and risk models (11). Recent activity is intense — 30 works in the last three years, with annual counts rising from 4 in 2021 to 19 in 2022, 15 in 2023 and 14 in 2025, earning the top activity score of 100. Three 2026 arXiv preprints reviewed by Evolve — quality-constrained routing over a fixed pool of quantized Mixture-of-Experts instances and two studies of pricing and inventory learning under contextual, censored demand — all place Zhang in the last author position, consistent with a senior, theory-driven supervisory role inside a compact six-person collaborator pool. The snapshot lists h-index 21 and i10-index 34, while the output and impact axes remain unscored for insufficient verified data. Biomedical topic clusters (glioma, ferroptosis) in the record likely reflect homonymous authors under a common name, so this profile should be read as name-resolved from public bibliographic data.

Collaboration pattern

Collaboration is compact: the recent three-year window lists 30 works with just 6 distinct coauthors, giving a collaboration score of 39. The recurring pool — Zean Han, Zezhen Ding, Jing Liang, Ruihan Lin, Zhenghong Huang and Hongfan Wu — reads like a stable advisee group, and on all three Evolve-reviewed 2026 preprints Zhang takes the last author position without a corresponding-author flag, a typical senior-supervisor signature. No large cross-institution network is visible in the provided data.

Research keywords

dynamic pricinginventory controlcensored demandcontextual banditsregret boundsminimax regretLLM inference servingMixture-of-Expertspost-training quantizationrequest routingqueueing theorylinear programming duality

Highlights

Quality-constrained routing over a fixed pool of quantized MoE instances
This 2026 preprint frames LLM request routing as a quality-constrained assignment over already-quantized Mixture-of-Experts instances, using linear programming duality and explicit quality-degradation budgets. Evolve's verdict praises the rigorously scoped claims and strong theory, while noting the 2.5% FWP gain measured on 88 prompts comes without variance estimates.
Learning to price and stock under contextual, censored demand
This revenue-management preprint claims minimax-optimal regret for joint dynamic pricing and inventory control when demand is contextual and censored. Evolve judges the theory credible but notes the proofs read as sketches and the experiments lack baselines; reproducibility readiness is rated medium.
Nonparametric contextual pricing and inventory learning with matching bounds
A companion paper declares MCK-UCB-based, minimax-optimal contextual pricing-inventory learning under censored sales with matching bounds. Evolve finds the proofs thorough but the experiments confined to synthetic data. The keyword set centers on censored demand, kernel UCB and minimax regret, reinforcing the revenue-management thread shared with the other two preprints.

Timeline

2008Earliest year in the OpenAlex countsByYear record, with 1 indexed work.
2012An early peak in the record with 5 indexed works.
2022Highest annual count in the record: 19 works, up from 4 in 2021.
202514 works indexed, part of a sustained high-output phase with 30 works in the recent three-year window.
2026Three Evolve-reviewed arXiv preprints appear: quality-constrained routing over quantized MoE instances, plus two pricing-and-inventory learning papers under contextual, censored demand.

Papers in the Evolve catalog

Frequent co-authors

Zean Han×2 Zezhen Ding×2 Jing Liang×1 Ruihan Lin×1 Zhenghong Huang×1 Hongfan 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 snapshot is from OpenAlex only; citation figures (h-index 21, i10-index 34) are quoted as provided and not independently verified.
Jiheng Zhang is a common name: the profile is name-resolved from public bibliographic data, and biomedical topic clusters (glioma, ferroptosis) in the record may belong to homonymous authors.
Output and impact axes are unscored due to insufficient verified data and are therefore described rather than estimated.
The reproducibility score rests on only 2 Evolve-reviewed papers (reproPapers 2, mean 55); all three reviewed preprints are 2026 arXiv postings and may still be under peer review.