Xiao Wang

Shanghai University of Finance and Economics · China

Unclaimed · AI-structured profileOpenAlex

Early-career researcher bridging log-concave sampling theory and privacy-aware memory for multi-user LLM agents.

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—

The output axis measures publication volume, venue quality, and lead-authorship. It cannot be scored yet: OpenAlex records h-index 0 and no i10-index, and only two 2026 arXiv preprints are in scope, with middle or last authorship.

Impact—

The impact axis would measure citations, citation velocity, and uptake of methods by other groups. No citation data is available in this snapshot, so it is left unscored rather than estimated from the two early preprints.

Focus0/100

OpenAlex's field distribution splits exactly 50/50 between LLM agent systems / natural language processing and theoretical statistics / sampling complexity, and all five indexed topics carry a count of one each, producing a focus score of 0. The two reviewed 2026 preprints sit at opposite ends: minimax query complexity for fixed-dimensional log-concave sampling versus a privacy-aware memory framework for multi-agent LLM systems.

Collaboration50/100

A collaboration score of 50 reflects a moderately connected early profile: 11 distinct coauthors appear across just two 2026 papers, with recurring partners such as Zachary Johnson, Min Chen, and Sadid A. Hasan. On both reviewed papers Xiao Wang holds a middle or last author position and is not the corresponding author, suggesting contribution within larger teams rather than independent leadership.

Activity32/100

The activity score of 32 is computed from a thin recent record: countsByYear shows exactly 2 works in 2026 and recent3yWorks equals 2, with no earlier years recorded. Both entries are arXiv preprints posted in the same year, indicating an initial burst of output on a very short bibliographic timeline rather than sustained multi-year activity.

Reproducibility55/100

Reproducibility scores 55, the mean Evolve readiness rating across 2 reviewed papers. Both carry a medium rating: the sampling paper's tight Θ(log(1+κ)+σ²/(με)) theory is considered strong, but its AI-co-built proofs need checking, while AIM's MUMBench benchmark suffers from synthetic circularity and missing baselines. Code and data availability remains the main gap for both.

Research summary

Xiao Wang is affiliated with Shanghai University of Finance and Economics in China. Evolve's catalog contains two 2026 arXiv preprints involving this researcher, and OpenAlex splits the profile's field distribution evenly (0.5/0.5) between LLM agent systems / natural language processing and theoretical statistics / sampling complexity — a dual-track, early-stage record. The first reviewed paper proves tight Θ(log(1+κ)+σ²/(με)) query complexity for sampling from fixed-dimensional log-concave distributions under stochastic gradient oracles; the author sits in the last position, and Evolve rates the theory strong while flagging that AI-co-built proofs need checking. The second, AIM, brings index-level privacy control to multi-agent multi-user LLM memory and introduces the MUMBench benchmark; Evolve finds the design promising but weakened by synthetic data circularity and missing baselines. Bibliometrically the profile is thin: h-index and i10-index are recorded as 0, countsByYear shows 2 works in 2026 (recent3yWorks = 2), and 11 distinct coauthors yield a collaboration score of 50. Both papers carry medium repro-readiness, averaging to a reproducibility score of 55. Because 'Xiao Wang' is a common name, this profile is name-resolved from public OpenAlex bibliographic data.

Collaboration pattern

Across the two reviewed 2026 preprints, Xiao Wang works inside sizeable teams: OpenAlex records 11 distinct coauthors, and the frequent-collaborator list includes Zachary Johnson, Somya Chatterjee, Min Chen, Xinyi Alice Li, Sadid A. Hasan, and Sulaiman Vesal. Authorship sits in middle and last positions on both papers, with no corresponding authorship — a pattern of contributing specialist expertise (sampling theory, memory-system privacy) within collaborative groups rather than leading them.

Research keywords

log-concave samplingstochastic gradient oracleminimax query complexitytotal variation distancerejection samplingcondition numberLLM agentsagentic memorymulti-user systemsprivacy and access controlmemory benchmarktopic modeling

Highlights

Tight query complexity for fixed-dimensional log-concave sampling under noisy gradients
In a 2026 arXiv preprint (last author), the theoretical-statistics line of this profile proves a tight Θ(log(1+κ)+σ²/(με)) minimax query complexity for sampling from log-concave distributions in fixed dimension with stochastic gradient oracles, measured in total variation distance; Evolve rates the theory strong but notes the AI-co-built proofs need independent checking.
AIM: index-level privacy control for multi-user LLM agent memory
The second 2026 preprint (middle author) proposes AIM, a privacy-aware interoperable memory framework for multi-agent multi-user LLM systems that adds index-level access control, and introduces the MUMBench memory benchmark; Evolve views the design as promising but flags synthetic data circularity and missing baselines as weaknesses.

Timeline

2026Posts 'Tight Sampling Complexity with stochastic gradient oracles in Fixed Dimensions' to arXiv as last author, proving tight Θ(log(1+κ)+σ²/(με)) query complexity for fixed-dimensional log-concave sampling.
2026Contributes as a middle author to 'AIM', a privacy-aware interoperable memory framework for multi-agent multi-user LLM systems with index-level access control and the MUMBench benchmark.
2026The OpenAlex snapshot records 2 works in the year, 11 distinct coauthors, and an even 0.5/0.5 field split between LLM agent systems and sampling complexity — an early-stage, dual-track profile.

Papers in the Evolve catalog

Frequent co-authors

Zachary Johnson×1 Nigel Boachie Kumankumah×1 Somya Chatterjee×1 Tejas Sathyamurthi×1 Min Chen×1 Xinyi Alice Li×1 Emily Morgan Gelchie×1 Jessica Lin×1 Sadid A. Hasan×1 Sulaiman Vesal×1 Weiming Ou×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.

'Xiao Wang' is a common name; this profile is name-resolved from public OpenAlex bibliographic data and may conflate distinct individuals.
Bibliometrics come from OpenAlex only; h-index and i10-index are recorded as 0, and the output and impact dimensions lack sufficient data to score.
Scope is limited to two 2026 arXiv preprints reviewed by Evolve; no peer-reviewed venue records are included.
The reproducibility score (55) averages Evolve reviews of just 2 papers, both rated medium readiness; it is not an assessment of a lab or codebase.