Alibaba Group researcher building scalable, verifiable environments for GUI and terminal LLM agents
2
papers here
2
reviewed by Evolve
0
first-author
0
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 score: the snapshot lacks a verified publication record for this author, so total output — paper counts, venues, and production rate over time — cannot be measured yet. This axis would track cumulative and annual research output.
Impact—
No score: citation counts, influential citations, and downstream uptake of the two cataloged environment papers are not available in this snapshot. This axis would measure how widely the work is read, cited, and adopted by other agent-environment efforts.
Focus0/100
The focus score is 0, driven by an even split of the cataloged record across two fields: machine learning / LLM agents (0.5) and machine learning for code / LLM agents (0.5). In practice the two streams are closely related — both target LLM agent environments, spanning GUI/computer-use settings and terminal settings — but the recorded field labels diverge enough that concentration registers as minimal on this snapshot.
Collaboration77/100
Collaboration scores 77, based on 46 distinct coauthors in the snapshot. The recurring collaborator list includes Que Shen, Shuai Bai, Jiayong Deng, Sicheng Fan, Yikun Fu, Chang Gao, Mianqiu Huang, Yizhen Jiang, Yuheng Jing, Dehui Kong, and Keliang Li. On both cataloged papers Zhang sits in a middle author position rather than first or corresponding, a pattern consistent with contribution to large industrial research teams building LLM agent environments.
Activity—
No score: the snapshot only covers two 2026 arXiv entries, with no year-by-year publication cadence to establish a trend. This axis would measure publishing frequency and continuity across recent years, which cannot be inferred from a single-year record.
Reproducibility60/100
Reproducibility averages 60 across the 2 papers with readiness data (reproMean 60, reproPapers 2), but the two papers diverge sharply: RecreationWorld is rated high for reproducibility readiness, with a benchmark and SFT pipeline described as rigorously evaluated, while Terminal-Universe is rated low — Evolve's review flags teacher-authored verifiers and the absence of a code release as reasons for caution before reuse.
Research summary
Zhenru Zhang is a researcher affiliated with Alibaba Group whose visible recent work centers on building scalable, verifiable environments for large language model (LLM) agents. The Evolve catalog holds two 2026 arXiv papers with Zhang in a middle author position. RecreationWorld introduces recreation environments recreated across five platforms, plus an SFT pipeline for hybrid GUI/coding agents; Evolve's review calls its evaluation design rigorous and rates reproducibility readiness high, while cautioning that reported transfer gains lack statistical controls. Terminal-Universe converts existing agent trajectories into scalable terminal training environments; ablations support a +11.9 gain on TB2.1, but teacher-authored verifiers and no code release put its reproducibility readiness at low. Both papers attack one bottleneck — the scarcity of verifiable environments for agents that operate graphical interfaces and command-line terminals. Bibliometric signals record 46 distinct coauthors (collaboration score 77), a reproducibility mean of 60 across two papers, and an even split between machine learning / LLM agents and machine learning for code / LLM agents; the focus score is 0, reflecting that two-field spread. Output, impact, and activity lack sufficient data and are left unscored. Because Zhenru Zhang is a common name, this profile is name-resolved from public bibliographic data.
Collaboration pattern
Zhang's recorded collaboration footprint is broad: 46 distinct coauthors (collaboration score 77), with a recurring core including Que Shen, Shuai Bai, Jiayong Deng, Sicheng Fan, Yikun Fu, Chang Gao, Mianqiu Huang, Yizhen Jiang, Yuheng Jing, Dehui Kong, and Keliang Li. On both cataloged papers Zhang holds a middle author position without corresponding status, consistent with contributing specialist work inside a large Alibaba Group research team rather than leading papers as sole lead.
Research keywords
Highlights
RecreationWorld: five-platform recreation benchmark for hybrid computer-use agents
Terminal-Universe: recycling agent trajectories into terminal training environments
One thesis across two papers: scaling verifiable agent environments
Both cataloged 2026 works attack the same bottleneck — the shortage of verifiable environments for training and evaluating LLM agents — from complementary directions: RecreationWorld recreates GUI platforms at scale, while Terminal-Universe synthesizes terminal tasks from replayed trajectories.
Timeline
- 2026Terminal-Universe appears on arXiv, converting agent trajectories into scalable terminal training environments (middle-author position, non-corresponding).
- 2026RecreationWorld appears on arXiv, introducing a five-platform recreation benchmark and an SFT pipeline for hybrid computer-use agents (middle-author position, non-corresponding).
- 2026Both cataloged papers land in the same year, establishing a visible focus on scalable, verifiable LLM agent environments spanning GUI and terminal settings.
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 (radar scores, coauthor counts, field distribution) come from OpenAlex-based bibliometric data only; no other citation source was consulted.
- No ORCID is linked and Zhenru Zhang is a common Chinese name; this profile is name-resolved from public bibliographic data and may involve record merging or splitting.
- The output, impact, and activity dimensions are unscored because the snapshot contains no citation counts, publication volume, or multi-year cadence data.
- Assessment rests on only two 2026 arXiv preprints in the Evolve catalog, both middle-author and non-corresponding; verdicts reflect Evolve's editorial review rather than journal peer review.