First author on LLM multi-agent memory co-evolution and diagnostic benchmarks for graph-enhanced MAS.
2
papers here
2
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) 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 aggregate output metrics (total publication count, venue record, career-wide productivity) are available yet — only the two 2026 Evolve-reviewed papers are on record. This axis would measure publication volume and consistency across venues and years.
Impact—
Citation counts, h-index and downstream influence are absent from the current data, so impact cannot be scored. This axis would measure how widely the two multi-agent papers are cited and adopted by follow-up work.
Focus0/100
Focus is recorded as 0 on the 0-100 scale, and the underlying field distribution calls for caution: only two papers are on record, splitting evenly (0.5 and 0.5) between 'LLM-based multi-agent systems' and its evaluation subfield. Both works sit squarely on graph-structured agent memory and graph-enhanced MAS, so thematically the profile looks highly concentrated, but the two-paper sample makes the score unstable.
Collaboration53/100
The collaboration score is 53, computed from 13 distinct coauthors across the reviewed work. The list repeats Xunkai Li, Zekai Chen and Rong-Hua Li, suggesting a stable core team, while Minghao An, Kaixiang Zhang, Yuxuan Ba, Ziheng Yi and Guangyuan He form a wider periphery. Kairui Yang appears as first author on both papers, indicating a leading rather than supporting role in these teams.
Activity—
No temporal data beyond two 2026-dated papers is available, so activity is unscored. This axis measures publication cadence and research continuity over time, which a single year of records cannot yet establish.
Reproducibility55/100
Reproducibility averages 55 across the two reviewed papers, both rated 'medium' readiness. Evolve's checks temper the headline numbers: MACE reports 81.11% versus SAGE's 78.97% over 8 benchmarks, but with single-run evaluations and sub-0.1pp margins, while OpenMAS-GCom's controlled-intervention evidence covers only five of its 29 datasets — rigorous design, incomplete empirical coverage.
Research summary
Kairui Yang is an early-stage researcher working at the core of LLM-based multi-agent systems, with two first-author arXiv papers from 2026 in the Evolve catalog. The first, MACE, proposes memory-agent co-evolution built on adaptive memory graphs, using experience reuse and feedback-driven adaptation over a graph-of-graphs structure; Evolve's review records that it tops 8 benchmarks at 81.11% versus 78.97% for the SAGE baseline, while cautioning that single-run evaluations, sub-0.1pp margins and uncited baselines leave the state-of-the-art claim unproven. The second, OpenMAS-GCom, is a diagnostic benchmark for graph-enhanced multi-agent systems, claiming controlled-intervention coverage across 29 datasets; Evolve judges the design rigorous but notes that intervention evidence currently spans only five datasets. Together the two papers split the field distribution evenly (0.5/0.5) between multi-agent methods and their evaluation, and both carry 'medium' reproducibility readiness, giving a reproducibility mean of 55. Collaboration data shows 13 distinct coauthors (collaboration score 53), with Xunkai Li, Zekai Chen and Rong-Hua Li recurring most often. No institution, country or ORCID is attached to this record, and the output, impact and activity axes remain unscored for lack of aggregate bibliometrics — so this should be read as a promising but early-stage profile anchored on two strong, critically-reviewed 2026 releases.
Collaboration pattern
Co-authorship clusters around one recurring group: Xunkai Li, Zekai Chen and Rong-Hua Li each appear twice in the collaboration list, alongside Minghao An, Kaixiang Zhang, Yuxuan Ba, Ziheng Yi and Guangyuan He — 13 distinct coauthors in total (score 53). Kairui Yang takes first authorship on both reviewed papers, consistent with a leading contributor inside a stable core team rather than a loose set of one-off partnerships.
Research keywords
Highlights
MACE memory-graph co-evolution leads 8 benchmarks at 81.11%
MACE's state-of-the-art claim flagged as unproven
OpenMAS-GCom: controlled-intervention diagnosis of graph-enhanced MAS
Timeline
- 2026Publishes MACE as first author on arXiv — memory-agent co-evolution with adaptive memory graphs; Evolve records 81.11% versus SAGE's 78.97% across 8 benchmarks.
- 2026Publishes OpenMAS-GCom as first author on arXiv — a diagnostic benchmark for graph-enhanced multi-agent systems claiming controlled interventions across 29 datasets.
- 2026Both papers enter the Evolve catalog rated 'medium' reproducibility readiness, setting a reproducibility mean of 55 across the two reviewed papers.
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.
- Yang is a common name; this profile is name-resolved from public bibliographic data (arXiv records in the Evolve catalog), with no institution, country or ORCID attached.
- Coverage is limited to two Evolve-reviewed 2026 arXiv papers; no citation counts, publication venues or career-wide publication history are included.
- The output, impact and activity radar axes are unscored because the underlying aggregate bibliometrics are missing (listed in dimMissing).
- The reproducibility score (mean 55) rests on only two papers rated 'medium', so treat it as indicative rather than stable.