Ying Shan: generative video world models and multimodal game-agent benchmarks, last author on both 2026 arXiv works
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-08) 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—
Data insufficient. This snapshot provides no verified publication count, so scholarly output cannot be scored. This axis would measure cumulative output — peer-reviewed papers, preprints, and their volume and distribution over time.
Impact—
Data insufficient. No citation counts, h-index, or downstream adoption metrics are present in the provided records. This axis would measure how widely and influentially this researcher's work is cited and built upon by others.
Focus0/100
Scores 0/100, reflecting an exactly even split in the raw field distribution: 0.5 machine learning (multimodal game agents and benchmarks) versus 0.5 computer vision (generative video world models). The two 2026 cataloged papers sit one per field, so current evidence shows two parallel research tracks rather than one concentrated agenda — breadth across generative modeling and game-agent evaluation, not depth in a single line.
Collaboration62/100
Scores 62/100. The bibliometric snapshot records 22 distinct coauthors, with 12 recurring collaborators — including Yiran Wang, Wangbo Yu and Xiaodong Cun — forming a stable core team. This moderately sized, lab-style network underpins both 2026 arXiv papers, one in each field, and positions Ying Shan as the last author on each.
Activity—
Data insufficient. With only two 2026 cataloged papers, recent-year publication frequency cannot be computed. This axis would measure ongoing research activity, such as annual output volume and submission cadence over time.
Reproducibility38/100
Scores 38/100, a mean over 2 reviewed papers. GameHorizon Suite rates medium readiness — large-scale data (5,000 hours, 21 games, 47 models) but VLM-generated ground truth and no online AAA evaluation. WorldCrafter rates low — its 47.6% revisit-consistency gain rests on a self-built benchmark. Independent replication is not yet demonstrated for either.
Research summary
Ying Shan (ORCID 0000-0001-7673-8325) is an early-stage profile documented through two 2026 arXiv preprints on which Ying Shan holds the last author position, neither as corresponding author. The bibliometric field distribution splits evenly at 0.5/0.5 between machine learning — multimodal game agents and benchmarks — and computer vision — generative video world models — so two parallel tracks define the current portfolio. The first produced GameHorizon Suite, a multi-horizon AAA gameplay dataset and benchmark spanning 5,000 hours, 21 games and 47 evaluated models; Evolve judges the scale solid while flagging VLM-generated ground truth and the absence of online AAA evaluation. The second produced WorldCrafter, a consistent video world model built on implicit 3D-aware memory, reporting a 47.6% revisit-consistency gain over 8 baselines with strong ablations, though the evidence rests on a self-built benchmark. Collaboration metrics record 22 distinct coauthors and 12 recurring collaborators, yielding a collaboration score of 62/100; reproducibility averages 38/100 across the two reviewed papers (medium and low readiness), and the even two-field split yields a focus score of 0/100. Output, impact and activity dimensions lack sufficient data, and no institution, country, citation record or peer-reviewed venue appears in this snapshot.
Collaboration pattern
The snapshot records 22 distinct coauthors, 12 of them recurring, including Yiran Wang, Wangbo Yu, Xingyilang Yin, Junfu Pu and Xiaodong Cun. This stable core backs both 2026 arXiv papers — a benchmark-engineering effort (GameHorizon Suite) and a model-research effort (WorldCrafter) — under a last-author placement on each. The 62/100 collaboration score reflects a moderately sized, lab-style team rather than a wide open-network pattern; institution data is unavailable, so no affiliation pattern can be inferred.
Research keywords
Highlights
GameHorizon Suite: 5,000 hours of AAA gameplay as a benchmark
WorldCrafter: implicit 3D-aware memory for consistent video worlds
Last-author role across both 2026 preprints
On both Evolve-reviewed papers Ying Shan appears in the last author position (neither as corresponding author), a placement commonly associated with supervisory or senior contribution; both works are arXiv preprints, so peer-review status is not yet established.
Timeline
- 2026WorldCrafter appears on arXiv: a consistent video world model with implicit 3D-aware memory, reporting a 47.6% revisit-consistency gain over 8 baselines.
- 2026GameHorizon Suite appears on arXiv: a multi-horizon AAA gameplay dataset and benchmark covering 5,000 hours, 21 games and 47 models.
- 2026Both cataloged works land in the same year, splitting the field distribution evenly between machine learning (game agents) and computer vision (video world models).
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.
- Bibliometric signals (dimension scores, 22 distinct coauthors, field distribution) come from a limited bibliometric snapshot; no citation counts, h-index or publication totals are included.
- 'Ying Shan' is a common name; this profile is name-resolved from public bibliographic data via ORCID 0000-0001-7673-8325, with no institution or country on record.
- Only 2 Evolve-reviewed papers, both 2026 arXiv preprints; peer-reviewed venues, funding and career history are outside this snapshot, so output, impact and activity dimensions are marked missing.
- Reproducibility reflects Evolve review of these 2 papers only (mean 38/100); broader code and data release behavior beyond them is unknown.