Zongyi Xu

Chongqing University of Posts and Telecommunications

Unclaimed · AI-structured profileORCID OpenAlex

Spatio-temporal ML researcher at CQUPT building self-supervised representations for urban regions, paths and trajectories

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

—

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-16) 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—

No publication-volume or productivity metrics are available in the current data, so this axis is unscored. It would measure the number, venues and recency of Xu's papers; at present only three 2026 arXiv entries exist as evidence.

Impact—

Citation counts, h-index and downstream adoption are absent from the input data, so impact cannot be scored yet. This axis would measure how widely Xu's representation-learning methods are cited and reused by other work.

Focus0/100

The radar scores focus at 0, and the underlying field distribution explains why: Xu's three cataloged papers split evenly — about one third each — across urban computing, intelligent transportation systems, and spatio-temporal data mining. Read closely, the three labels describe one coherent spatio-temporal machine-learning agenda — urban regions, paths, trajectories — rather than scattered interests; the score reflects label-level breadth, not lack of direction.

Collaboration50/100

With a collaboration score of 50, Xu's recorded network contains 11 distinct co-authors, concentrated in one recurring cluster: Sean Bin Yang, Jilin Hu, Hua Lu, Bin Yang, Christian S. Jensen, Ying Sun, Kristian Torp, Hao Miao, Xiangmeng Wang, Tung Kieu and Torben Bach Pedersen. On all three cataloged papers Xu sits as a middle, non-corresponding author, indicating contribution within established teams rather than independent leadership so far.

Activity—

Beyond three papers all dated 2026, no publication-frequency or year-over-year volume data is available, so activity is unscored. This axis would measure publishing cadence and recent engagement over time.

Reproducibility70/100

Evolve's reproducibility assessment covers three papers and averages 70/100. DGCPath rates high readiness, while CURE and REFINE are medium; reviews credit broad experiments (up to ten baselines, three to four datasets) but note thin significance testing and conditional theory. This is the strongest quantified signal in Xu's current profile.

Research summary

Zongyi Xu is a machine-learning researcher at Chongqing University of Posts and Telecommunications (ORCID 0000-0002-8267-6834) whose cataloged work centres on self-supervised representation learning for spatio-temporal data. Evolve's catalog lists three 2026 arXiv papers, spread evenly (about one third each) across three closely related subfields: urban computing, intelligent transportation systems, and spatio-temporal data mining. The first, CURE, adds confounder-aware shared-component residualization to multi-view urban region embeddings, reporting strong gains on three cities. The second, DGCPath, combines diffusion-generated views with Jensen-Shannon distributional contrastive learning for path representations, outperforming ten baselines across three trajectory datasets. The third, REFINE — a KDD'26 extended paper — unifies contrastive and generative trajectory representation learning through closed-loop transcription, with broad wins on four datasets. Across the three papers, Evolve's reproducibility assessment averages 70/100 — high for DGCPath, medium for CURE and REFINE — although reviews flag thin statistical testing and, for CURE, limited causal rigor. Xu appears as a middle, non-corresponding author on all three papers, working within an 11-person co-author network that includes Sean Bin Yang, Christian S. Jensen, Hua Lu, Bin Yang and Hao Miao. Output, impact and activity metrics are not yet available for this profile.

Collaboration pattern

Xu's recorded collaboration pattern is team-embedded: on all three cataloged papers Xu is a middle, non-corresponding author. The network of 11 distinct co-authors is dominated by one recurring cluster — Sean Bin Yang, Jilin Hu, Hua Lu, Bin Yang, Christian S. Jensen, Ying Sun, Kristian Torp, Hao Miao, Xiangmeng Wang, Tung Kieu and Torben Bach Pedersen — indicating work within an established spatio-temporal data research group rather than a spread of independent partnerships.

Research keywords

urban region representation learningtrajectory representation learningpath representation learningself-supervised learningcontrastive learningdiffusion modelsmulti-view learningconfounder-aware learninggraph neural networksmasked reconstructionspatio-temporal data miningdeconfounding

Highlights

Confounder-aware multi-view urban region embeddings
CURE adds heuristic shared-component residualization to multi-view urban embeddings and reports strong gains across three cities; Evolve's review finds the deconfounding lacks causal rigor and statistical support.
Distribution-aware contrastive path representations
DGCPath pairs diffusion-generated views with Jensen-Shannon distributional contrastive learning, outperforming ten baselines on three trajectory datasets; Evolve rates its reproducibility high, while judging novelty incremental and significance tests absent.
Closed-loop transcription for trajectory representation
The KDD'26 extended paper unifies contrastive and generative trajectory representation learning via closed-loop transcription, winning broadly across four datasets; Evolve's verdict notes conditional theory and thin statistics.

Timeline

2026Publishes CURE on arXiv: confounder-aware multi-view urban region representation learning, evaluated on three cities.
2026Publishes DGCPath (extended version): diffusion-augmented, distribution-aware contrastive path representation learning, validated on three trajectory datasets against ten baselines.
2026Publishes REFINE (KDD'26 extended version): closed-loop transcription unifying contrastive and generative trajectory representation learning across four datasets.
2026All three cataloged works appear with Xu as a middle, non-corresponding author within an 11-person co-author network.

Papers in the Evolve catalog

Frequent co-authors

Sean Bin Yang×3 Jilin Hu×3 Hua Lu×3 Bin Yang×3 Christian S. Jensen×2 Ying Sun×2 Kristian Torp×2 Hao Miao×1 Xiangmeng Wang×1 Tung Kieu×1 Torben Bach Pedersen×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 scores and co-author counts are derived from OpenAlex-based data only; no citation, h-index or venue-level metrics were available.
Output, impact and activity axes are unscored because the input contains no publication-volume or citation data — only three 2026 arXiv papers.
All cataloged papers are arXiv versions with Xu as a middle, non-corresponding author; first-author or corresponding-author work is not represented in this profile.
'Zongyi Xu' is a common Chinese name; this profile is name-resolved from public bibliographic data and anchored to ORCID 0000-0002-8267-6834 — verify identity before citing.