Yi Zhang

University of Technology Sydney · Australia

Unclaimed · AI-structured profileOpenAlex

Yi Zhang (UTS): 3D vision meets embodied world models — from Gaussian splatting to cross-embodiment simulators

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—

The output axis measures publication volume, venue tier and authorship role over time. It is not scored here: the snapshot records only 2 works (both 2026 arXiv preprints) with no venue history, so any score would be unreliable.

Impact—

The impact axis tracks citations, h-index and i10-index. These read 0 in the snapshot, which — given no linked ORCID and a very common name — likely reflects unresolved identity, so no impact score is assigned.

Focus0/100

Focus scores topical concentration, and this record sits at 0: the field distribution splits exactly 50/50 between robotics/embodied AI (world models) and computer vision (3D scene representation / novel view synthesis), and the five recorded topics (e.g. Reinforcement Learning in Robotics, 3D Shape Modeling and Analysis) each appear only once. Activity is thus spread evenly across two adjacent but distinct areas rather than concentrated in one.

Collaboration51/100

Collaboration scores network breadth and role, and this record is mid-range at 51. The snapshot counts 12 distinct coauthors; both catalog papers are large-team efforts where Yi Zhang is a middle author on Pelican-Sim 1.0 and last author on the Gaussian Splatting study, though neither lists them as corresponding author. Frequent coauthors include Shilong Zou, Shilin Zhang, Yingji Zhang, Yuhang Huang, Xin Cao and Jian Tang.

Activity32/100

Activity scores recent publication cadence, scoring 32 here. The counts-by-year record contains a single year, 2026, with 2 works, and recent3yWorks is likewise 2 — both arXiv preprints. That is a thin recent record, and given the unresolved identity it likely understates true output; on the evidence captured, recent activity is low and concentrated entirely in 2026.

Reproducibility55/100

Reproducibility averages Evolve's readiness assessments across 2 scored papers, giving a mean of 55 — medium readiness on both. The reviews flag concrete gaps: the Gaussian Splatting study needs random seeds, more scenes and released code, while Pelican-Sim 1.0 reports strong ablations but no error bars and possible train-eval overlap. The middling score reflects solid evaluation design paired with unreleased experimental detail.

Research summary

Yi Zhang is affiliated with the University of Technology Sydney and works at the intersection of computer vision and embodied AI, with an OpenAlex field split of 50% robotics/embodied AI (world models) and 50% computer vision (3D scene representation / novel view synthesis). The Evolve catalog holds two 2026 arXiv papers. 'Is Gaussian Splatting Becoming Neural Again?' is a taxonomy and controlled study of learned parameterization for 3D Gaussian Splatting, with Yi Zhang as last author; Evolve's verdict credits the finding that shared appearance parameterization helps while geometry decoding does not, while flagging missing seeds, limited scenes and no code (medium reproducibility readiness). 'Pelican-Sim 1.0' is a general world-model simulator for embodied intelligence combining mixture-of-experts and diffusion-transformer components for action-conditioned video generation and cross-embodiment robot manipulation, with Yi Zhang as a middle author; strong ablations are tempered by absent error bars and possible train-eval overlap. Recorded metrics are thin and likely deflated: h-index and i10-index read 0, the output and impact axes are unscored, activity scores 32 with only 2 works counted (both 2026), and 12 distinct coauthors yield a mid-range collaboration score of 51; reproducibility averages 55 across the 2 scored papers. With no linked ORCID and a very common name, this is an early-stage, name-resolved profile built from public bibliographic data plus Evolve's own reviews.

Collaboration pattern

Both cataloged papers are large-team arXiv efforts: Yi Zhang is last author on the Gaussian Splatting taxonomy study and a middle author on Pelican-Sim 1.0, though neither lists them as corresponding author. The OpenAlex snapshot counts 12 distinct coauthors (collaboration score 51, mid-range), with recurring partners including Shilong Zou, Shilin Zhang, Yingji Zhang, Yuhang Huang, Xin Cao, Jian Tang, Yong Dai and Xiaozhu Ju — a common-surname cluster that complicates author disambiguation.

Research keywords

3D Gaussian SplattingNeural Radiance FieldsLearned ParameterizationNovel View SynthesisTaxonomyAmortized InferenceWorld ModelsAction-Conditioned Video GenerationCross-Embodiment LearningRobot ManipulationMixture of ExpertsDiffusion Transformer

Highlights

Selective neuralization for 3D Gaussian Splatting
A taxonomy and controlled study of learned parameterization finds that neural components help when appearance is shared across views, while decoding geometry from learned features does not; Evolve rates reproducibility medium, noting missing seeds, limited scenes and no released code.
Pelican-Sim 1.0: a general world-model simulator for embodied AI
An action-conditioned video generation world model built with mixture-of-experts and diffusion-transformer components targets cross-embodiment robot manipulation, with large claimed gains and strong ablations; Evolve tempers the verdict over missing error bars and possible train-eval overlap.
Even split between 3D vision and embodied world models
The two cataloged 2026 works divide evenly (50/50) between computer vision (3D scene representation / novel view synthesis) and robotics/embodied AI (world models), with 12 distinct coauthors and team roles as last author on one paper and middle author on the other.

Timeline

2026Co-authors 'Is Gaussian Splatting Becoming Neural Again?' (arXiv), a taxonomy and controlled study of learned parameterization, as last author.
2026Co-authors 'Pelican-Sim 1.0' (arXiv), a general world model simulator for embodied intelligence, as a middle author.
2026OpenAlex snapshot records 2 works in 2026, 12 distinct coauthors, and a field split evenly between robotics/embodied AI (world models) and computer vision (3D scene representation).

Papers in the Evolve catalog

Frequent co-authors

Shilong Zou×1 Shilin Zhang×1 Yingji Zhang×1 Yuhang Huang×1 Zeyuan Ding×1 Han Dong×1 Junwei Liao×1 Yong Dai×1 Jian Tang×1 Xiaozhu Ju×1 YuanHang Wang×1 Xin Cao×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 counts come from OpenAlex only; no ORCID is linked, and 'Yi Zhang' is a very common name, so this profile is name-resolved from public bibliographic data and may merge or split identities.
Output and impact axes are unscored: h-index and i10-index read 0 in the snapshot, consistent with an unresolved identity rather than measured zero impact.
Only two 2026 arXiv preprints are in the Evolve catalog; peer-reviewed venue status, funding and awards are not captured in the input data.
Counts-by-year shows activity only in 2026; earlier or parallel output under other name variants is not represented.