Researcher at Palo Alto University working across world models and neural rendering, with sharply rising recent output
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
10
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—
Output is not yet scored: this snapshot lacks reliable per-publication records (venues, publication types, authorship roles) needed to size total output honestly. Once ingested, the axis would track annual works, first/corresponding-author share and venue quality.
Impact—
Impact is unscored because citation-side data is absent from this snapshot. The axis would aggregate citation counts, citing works and field-weighted influence to show how often and how widely other researchers build on these papers.
Focus0/100
The focus score is 0, the floor of the scale, reflecting an almost perfectly split portfolio: fieldDist assigns 0.5 to machine learning (video generation / world models) and 0.5 to computer graphics (neural rendering). Topic counts confirm the spread — generative adversarial networks and image synthesis (34), multimodal machine learning applications (18), computer graphics and visualization (15), digital media and visual art (13), and advanced neural network applications (10). This reads as deliberate breadth across two fields rather than a single-line agenda.
Collaboration71/100
Collaboration scores 71, supported by 34 distinct co-authors across the recent three-year window and its 31 works. Twelve partners recur, including Haotian Zhang, Fengyuan Yu, Dezhi Luo, Haoran Sun, Zehong Zhao and Qingying Gao. In both catalog papers Zhang holds a middle author position without corresponding status, indicating contribution within larger teams rather than sole or senior-led authorship.
Activity100/100
Activity scores 100, the maximum. countsByYear shows an accelerating curve: 1 work in 2017, 1 in 2018, 3 in 2020, 4 in 2021, then 3 in 2024, 11 in 2025 and 17 in 2026. The 2024–2026 window sums to 31 works, matching the recent-three-year total of 31, and both Evolve-reviewed papers are 2026 arXiv preprints — consistent with an intense, current publishing pace.
Reproducibility78/100
Reproducibility scores 78, computed as a mean of 78 across 2 assessed papers (reproPapers 2). Readiness differs by project: the world-model benchmark paper is rated high — 150 Blender generators and 1.5M samples are concretely specified — while RenderFormer-V2 is rated medium, its out-of-distribution gains shown only qualitatively.
Research summary
Lvmin Zhang (ORCID 0000-0003-3503-5791) is profiled here from public bibliographic data, name-resolved via OpenAlex; the listed affiliation is Palo Alto University, US. The snapshot shows a two-track record: the field distribution splits evenly, 0.5/0.5, between machine learning (video generation / world models) and computer graphics (neural rendering). Publishing pace is accelerating — indexed works rose from 3 in 2024 to 11 in 2025 and 17 in 2026, summing to 31 works over the recent three years, with h-index 10 and i10-index 10. Collaboration is broad: 34 distinct co-authors recently and twelve recurring partners, including Haotian Zhang, Fengyuan Yu and Qingying Gao; both catalog papers place Zhang in a middle, non-corresponding author position. Two 2026 preprints anchor the record. Training Object Permanence in World Models tests object permanence with 150 Blender generators and 1.5M samples — the fine-tuned model tops its class, though Evolve flags train-exam overlap. RenderFormer-V2 claims transformer-based global illumination of 100k+ heterogeneous primitives without per-scene training, with out-of-distribution gains shown only qualitatively. Reproducibility scores 78, a mean across 2 assessed papers. Citation-based output and impact dimensions are not yet scored.
Collaboration pattern
With 34 distinct co-authors across 31 recent works, Zhang operates inside sizeable, changing teams. Twelve partners recur, led by Haotian Zhang, Fengyuan Yu, Dezhi Luo, Haoran Sun, Zehong Zhao and Qingying Gao; the remaining names cluster around single project cycles. Both catalog papers place Zhang in a middle author position with no corresponding role, so the pattern reads as a high-bandwidth collaborator within group efforts rather than a solo or senior-led line.
Research keywords
Highlights
WROP benchmark stress-tests object permanence at scale
Transformer-based global illumination without per-scene training
Benchmark-first, reproducibility-minded output
Timeline
- 2017Earliest indexed works appear in the OpenAlex snapshot, with 1 work counted for 2017 and 1 for 2018 — a thin early record.
- 2021A modest early peak: 4 indexed works, the highest annual count before 2024.
- 2025Output accelerates — 11 indexed works, the first double-digit year in the record.
- 202617 indexed works so far; both Evolve-reviewed papers appear this year — the object-permanence benchmark and RenderFormer-V2.
Papers in the Evolve catalog
Frequent co-authors
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.
- Bibliometrics, dimension scores and topic counts come from an OpenAlex-based snapshot; the output and impact axes are unscored because citation-side data is missing.
- Identity is name-resolved from public bibliographic data under ORCID 0000-0003-3503-5791; the name appears across multiple authors, so the affiliation should be independently confirmed.
- The reproducibility score (78) rests on only 2 assessed papers, an early signal rather than a stable estimate.
- Timeline entries are derived from indexed work counts by year and the two catalog-reviewed papers, not from verified career milestones.