AI researcher whose first-author work compiles winning game players from bare interaction via LLM coding agents
1
papers here
1
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-21) 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—
Insufficient data: no publication counts, venue record, or output-volume metrics are available for this profile. This axis would measure the scale and consistency of the researcher's publishing record over time, and cannot yet be scored.
Impact—
Insufficient data: no citation counts, h-index, or downstream-usage metrics are available. This axis would measure the work's influence on subsequent research and practice; with a single recent 2026 paper, it is too early to score.
Focus100/100
The focus axis scores 100, the maximum. The underlying raw data shows all cataloged output — currently one 2026 arXiv paper — sits in a single field, 'artificial intelligence (LLM agents, game-playing evaluation).' With a one-paper sample, this score is better read as extreme concentration than as deliberate strategy: it confirms thematic unity today but cannot yet demonstrate a sustained, self-directed research agenda across multiple years.
Collaboration14/100
The collaboration axis scores 14, the lowest reported score. The raw count behind it is one distinct coauthor — Haonan Huang — across the cataloged record, and the single 2026 paper lists Xiao as first author (not corresponding). This is a strongly solo-led, two-person working pattern. With a sample of one paper, the score signals a small current network rather than a confirmed long-term aversion to collaboration.
Activity—
Insufficient data: the record spans a single year (2026), so no publication cadence or trend can be established. This axis would measure publishing rhythm and research momentum across consecutive years, which cannot yet be assessed.
Reproducibility55/100
The reproducibility axis scores 55, computed as a mean over exactly one reviewed paper (reproPapers: 1), whose own readiness is rated medium. Evolve's review adds texture: the StarCraft II evidence is strong, but session counts are tiny and the Civilization experiments were incomplete at the paper's freeze — conditions that typically limit independent replication. A one-paper mean is a real signal, but a fragile one.
Research summary
Joey Xiao is an early-stage AI researcher whose cataloged work centers on large language model (LLM) agents — specifically, using frontier general-purpose coding agents to synthesize game-playing policies directly from raw interaction. The Evolve catalog currently holds one reviewed paper, a 2026 arXiv preprint on which Xiao is first author, arguing that such agents can 'compile' winning players for games ranging from Flappy Bird to StarCraft II and Civilization within a single bare interaction session. Evolve's review rates the StarCraft II evidence as strong while flagging that session counts are tiny and the Civilization experiments were incomplete at the paper's freeze; the paper's reproducibility readiness is rated medium (reproducibility mean 55 over one paper). Bibliometrically, the profile is extremely concentrated: the focus score is 100, reflecting a single cataloged field (artificial intelligence — LLM agents, game-playing evaluation), and the collaboration score is 14, with exactly one distinct coauthor, Haonan Huang. No reliable output, impact, or activity metrics are available yet, so this profile should be read as a snapshot of one high-intensity first-author paper rather than an established research record.
Collaboration pattern
The cataloged record shows a strongly solo-led pattern: one 2026 first-authored paper and exactly one distinct coauthor, Haonan Huang, yielding a low collaboration score of 14. There is no evidence yet of advising networks, cross-institution teams, or repeat collaborations. With a sample of one paper, this is best read as an early two-person collaboration footprint rather than a settled working style; watch whether future cataloged papers add coauthors.
Research keywords
Highlights
First-authored arXiv study on 'compiled agency'
Strong StarCraft II evidence, open reproducibility questions
Timeline
- 2026First cataloged paper: a first-authored arXiv preprint on compiled agency, building winning game players from bare interaction from Flappy Bird to StarCraft II and Civilization.
- 2026Evolve review published: StarCraft II evidence rated strong, with tiny session counts and incomplete Civilization exams flagged; reproducibility readiness set to medium.
- 2026Bibliometric snapshot: focus score 100 in a single field, collaboration score 14 with one distinct coauthor (Haonan Huang), and a reproducibility mean of 55 across one paper.
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 scores (focus, collaboration, reproducibility) are computed solely from the Evolve catalog's bibliographic snapshot, whose coverage may lag external databases.
- The record contains a single 2026 paper; output, impact, and activity axes are unscored because no publication-volume or citation data was provided.
- No institution, country, or ORCID is on record; this profile is name-resolved from public bibliographic data, and 'Joey Xiao' may refer to more than one researcher.
- Paper assessments reflect Evolve's review at its snapshot date, and the authors themselves state that the Civilization experiments were incomplete at the paper's freeze.