Minghao An
—
Machine learning researcher working on graph-structured LLM multi-agent systems: topology, memory, and evaluation
3
papers here
3
reviewed by Evolve
0
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—
Evolve has no publication-count or venue-volume metrics for this researcher yet; the output axis would measure total papers, venues, and authorship roles over time. The three reviewed arXiv preprints are insufficient to score it.
Impact—
No citation counts, h-index, or citation-based impact data are available in the input. The impact axis would measure citations, uptake, and influence of the work; it is left unscored rather than estimated.
Focus8/100
Focus scores 8 out of 100, reflecting an exceptionally narrow research footprint. The field distribution sits entirely inside LLM-based multi-agent systems: about 67% on MAS construction (MAGIC, MACE) and about 33% on MAS evaluation (OpenMAS-GCom). All three 2026 papers share graph-based representations of agent systems, so the low score captures deep concentration in one niche rather than scattered output.
Collaboration54/100
Collaboration scores 54, computed from 14 distinct coauthors across the reviewed papers. Recurring names include Kairui Yang, Xunkai Li, Zekai Chen, and Rong-Hua Li, alongside Kaixiang Zhang, Yuxuan Ba, Ziheng Yi, and Zhanke Liu, suggesting a stable core team. On all three papers this researcher sits in a middle authorship position and never as corresponding author, indicating a contributor rather than a lead role so far.
Activity—
Activity — publication cadence, recency, and sustained output over time — is not yet computed for this profile. All three reviewed papers are dated 2026, which hints at recent activity but is too small a sample to score; the axis remains unscored.
Reproducibility55/100
Reproducibility averages 55 across three assessed papers, each rated medium readiness. Evolve's reviews explain the middling score: MAGIC's eight-benchmark lead rests on single-run results; MACE's 81.11% versus SAGE 78.97% involves sub-0.1pp margins and uncited baselines; OpenMAS-GCom's intervention evidence covers only five of its 29 claimed datasets. Methods are clearly described, but reported results still need independent replication.
Research summary
Minghao An is a machine learning researcher whose reviewed work centers on LLM-based multi-agent systems and, more specifically, on graph-structured approaches to building and evaluating them. The Evolve catalog holds three 2026 arXiv papers, all with this researcher in a middle authorship position and none as corresponding author. MAGIC learns per-role mixed-granularity agent topologies through incremental construction and dense-reward reinforcement learning, ranking first across eight benchmarks; MACE introduces memory–agent co-evolution over adaptive memory graphs, reported at 81.11% versus SAGE's 78.97% on eight benchmarks; OpenMAS-GCom proposes a diagnostic benchmark using controlled interventions to probe graph-enhanced MAS components across a claimed 29 datasets. Evolve's reviews are broadly positive on the ideas but flag evaluation caution: single-run results, sub-0.1pp margins, and intervention evidence spanning only five datasets. Bibliometric signals reinforce that picture. The field distribution splits roughly two-thirds LLM-based multi-agent systems and one-third their evaluation, giving an unusually narrow focus score of 8. Reproducibility averages 55 across three assessed papers with medium readiness, and collaboration scores 54 with 14 distinct coauthors. Because output, impact, and activity metrics are not yet available, this profile should be read as early-stage and name-resolved from public bibliographic data.
Collaboration pattern
Across three 2026 arXiv papers, Minghao An works with 14 distinct coauthors, with Kairui Yang, Xunkai Li, Zekai Chen, and Rong-Hua Li recurring most often, alongside Kaixiang Zhang, Yuxuan Ba, Ziheng Yi, and Zhanke Liu. This points to a stable, likely single-group collaboration pattern around graph-based LLM multi-agent systems. Consistent middle authorship with no corresponding role suggests a contributing team-member position rather than project leadership.
Research keywords
Highlights
MAGIC: RL-learned mixed-granularity agent graphs rank first on eight benchmarks
MACE: memory–agent co-evolution over adaptive memory graphs
OpenMAS-GCom: a controlled-intervention diagnostic benchmark for graph-enhanced MAS
Timeline
- 2026Publishes MAGIC, learning mixed-granularity agent graph topologies with dense-reward reinforcement learning; ranked first on eight benchmarks.
- 2026Publishes MACE, pairing agent learning with adaptive memory-graph co-evolution; reports 81.11% versus SAGE's 78.97%.
- 2026Contributes to OpenMAS-GCom, a controlled-intervention benchmark for diagnosing graph-enhanced multi-agent systems.
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.
- Radar scores are computed from public bibliometric data only; no institution, country, or ORCID is attached to this profile.
- Coverage is limited to three Evolve-reviewed 2026 arXiv preprints and may not reflect the full publication record.
- Output, impact, and activity axes lack underlying data and are left unscored rather than estimated.
- 'Minghao An' is name-resolved from public bibliographic data; identity disambiguation beyond that is not verified.