LLM agent research at Fudan University: verifiable computer-use environments and efficiency-oriented GUI agent systems
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) 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—
Not yet scored: this axis measures publication volume and productivity — total papers, first/corresponding-author share, and yearly output. The current snapshot does not include per-author publication counts, so no output figure can be reported.
Impact—
Not yet scored: this axis would measure citation-based influence, such as citation counts and the h-index. The input snapshot contains no citation data for Jiakang Yuan, so no impact estimate is offered here.
Focus0/100
The computed focus score is 0, driven by an even field split in the raw data: 50% machine learning / LLM agents and 50% artificial intelligence (NLP / GUI agents). Read substantively, however, the two labelled fields point at the same object of study — both catalogued 2026 papers target GUI and computer-use agents, one building verifiable environments and one surveying efficiency — so the topical core appears narrower than the split score implies.
Collaboration76/100
Collaboration scores 76/100, backed by a raw count of 44 distinct coauthors in the bibliographic record. Both Evolve-catalogued papers place Jiakang Yuan in a middle author position on large team rosters, a pattern typical of benchmark-building projects; the frequent-coauthor list includes Shuai Bai, Jiayong Deng, Sicheng Fan, Chang Gao, and Mianqiu Huang, among others. The breadth suggests heavy team-based work rather than small-group or solo authorship.
Activity—
Not yet scored: this axis tracks publication cadence — papers per year and recency of output. Only two 2026 arXiv entries are present in the snapshot, too few to compute a trend, so activity is left unscored.
Reproducibility60/100
Reproducibility scores 60/100, computed as a mean over the 2 papers Evolve has reviewed (reproMean 60). The split is concrete: RecreationWorld is rated high reproducibility readiness, consistent with its release of a five-platform recreation benchmark and SFT pipeline, while the efficient-GUI-agents systems survey is rated low, as surveys aggregate unnormalized per-paper numbers. Released artifacts and benchmarks anchor the score; secondary synthesis work pulls it down.
Research summary
Jiakang Yuan is a researcher at Fudan University working at the intersection of large language model agents and computer-use AI. This profile is name-resolved from public bibliographic data and covers the papers Evolve has reviewed to date: two 2026 arXiv releases in which Yuan appears as a middle author on large team rosters. The first, RecreationWorld, introduces a five-platform recreation benchmark with verifiable rewards plus an SFT pipeline for hybrid GUI/coding agents; Evolve judges its evaluation design rigorous while noting that reported transfer gains lack statistical controls. The second is a systems survey of efficient GUI agents organized around observation, memory, action, and runtime optimization; its per-paper reported-versus-unreported cost tables are flagged as real value, though cross-paper numbers remain unnormalized. Bibliometric signals reflect this team-based mode: 44 distinct coauthors and a collaboration score of 76/100, with recurring collaborators including Shuai Bai, Chang Gao, and Mianqiu Huang. The field distribution splits evenly (50/50) between machine learning / LLM agents and artificial intelligence (NLP / GUI agents), yielding a computed focus score of 0 even though both works target the same agent-systems problem space. Reproducibility scores 60/100 across the two reviewed papers, anchored by the high-reproducibility benchmark release. Citation, output-volume, and activity metrics are not yet available for this profile.
Collaboration pattern
Jiakang Yuan's catalogued work is team-based: both 2026 papers carry large author lists with Yuan in a middle position, and the record shows 44 distinct coauthors (collaboration score 76/100). Recurring collaborators in the snapshot include Shuai Bai, Jiayong Deng, Sicheng Fan, Chang Gao, Mianqiu Huang, Yizhen Jiang, Dehui Kong, Keliang Li, and Ning Li. The pattern suggests contribution to shared benchmark-and-systems projects rather than solo or small-group research.
Research keywords
Highlights
RecreationWorld: verifiable five-platform environments for hybrid computer-use agents
Systems survey maps efficiency levers for GUI agents
Contrasting reproducibility readiness across outputs
Reproducibility averages 60/100 over the two reviewed papers: the RecreationWorld benchmark release rates high readiness, while the systems survey rates low because it aggregates unnormalized per-paper numbers.
Timeline
- 2026Co-authored RecreationWorld, a five-platform recreation benchmark with verifiable rewards and an SFT pipeline for hybrid GUI/coding agents (arXiv, middle author).
- 2026Co-authored a systems survey of efficient GUI agents organized around observation, memory, action, and runtime optimization (arXiv, middle author).
- 2026Bibliographic record for the year's two catalogued outputs shows 44 distinct coauthors, a collaboration score of 76/100, and an even 50/50 field split across the two agent-related fields.
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 are computed from public bibliographic (OpenAlex-sourced) data; citation counts and h-index are not part of this snapshot.
- Profile is name-resolved from name plus institution (no ORCID provided); the name Jiakang Yuan may correspond to multiple researchers.
- Only the 2 papers reviewed by Evolve are covered; the catalog may not represent the researcher's full output.
- Both catalogued papers are 2026 arXiv preprints; peer-review status and venue acceptance are not recorded.