NLP researcher working on open Armenian language-model ecosystems and model-pool selection in LLM multi-agent systems
2
papers here
2
reviewed by Evolve
1
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—
Output data is not yet available: this axis would measure the volume, venue mix, and consistency of the researcher's publications over time. With no publication counts in the current record, no output score is estimated.
Impact—
Impact data is not yet available: this axis would measure citation-based influence such as total citation counts and h-index. The current record contains no citation statistics, so no impact score is estimated here.
Focus0/100
The available field profile splits evenly, 50% natural language processing for low-resource languages and 50% machine learning work on LLM multi-agent systems, which yields a focus score of 0 — no single dominant field. Both catalogued 2026 papers sit one in each area, so the portfolio currently spans two distinct research directions rather than concentrating in one.
Collaboration51/100
The catalogued work involves 12 distinct coauthors, and the frequent-collaborator list includes Khatun Avetisyan, Hayk Shahsuvaryan, Jiacheng Xu, and Sara Vera Marjanović, among others. The collaboration score of 51 falls in the middle range, consistent with a stable recurring team rather than either a very wide or a very thin coauthor network.
Activity—
Activity data is not yet available: this axis would measure publishing cadence and recent engagement across successive years. Only two 2026 papers are on record, too thin a base to score ongoing research activity.
Reproducibility78/100
Reproducibility is the strongest measured axis: the two catalogued papers average 78 on Evolve's reproducibility assessment (reproMean 78 across 2 papers). The first-author Armenian ecosystem paper is rated high readiness — described as fully open and reproducible with excellent data hygiene — while the multi-agent model-pool study is rated medium, its headline curves noted to lack variance estimates.
Research summary
Erik Arakelyan is a researcher working across two complementary lines of language-models research: open NLP for low-resource languages and the design of LLM-based multi-agent systems. The profile is name-resolved from public bibliographic data; Evolve's catalog currently holds two of the researcher's 2026 arXiv papers. The first, led as first author, is 'From Zero to Hero: An Open LLM Ecosystem for Armenian' — reviewed as the first fully open, reproducible Armenian LLM ecosystem, with excellent data hygiene, though single-run training and MT-derived benchmarks temper the headline claims. The second, a middle-authorship contribution, is 'Mo' Models, Mo' Problems', a systematic 23-model study arguing that growing heterogeneous model pools can degrade multi-agent system performance below its best single model — deemed credible overall, with headline curves lacking variance estimates. The field distribution is an even 50/50 split between the two areas, and the work draws on 12 distinct coauthors with a recurring core team. Reproducibility is the strongest measured signal, averaging 78 across the two assessed papers (high readiness for the Armenian ecosystem release, medium for the model-pool study). Output, impact, and activity metrics are not yet available in the underlying data and are left unscored, so conclusions rest on a small, very recent catalog.
Collaboration pattern
Erik Arakelyan's catalogued work draws on 12 distinct coauthors, yielding a mid-range collaboration score of 51. The frequent-collaborator list combines a stable recurring core cluster — Khatun Avetisyan, Meri Davtyan, Heghine Grigoryan, Nane Khachatryan, Hayk Shahsuvaryan, Henrik Sergoyan, Vahan Martirosyan, Grigor Nalbandyan, Evelina Bakhaturina — with collaborators including Sara Vera Marjanović, Jiacheng Xu, and Aleksandr Laptev, suggesting team-based projects plus selective external partnerships.
Research keywords
Highlights
First fully open, reproducible LLM ecosystem for Armenian
Data hygiene and contamination awareness in low-resource corpus work
Systematic 23-model study of model-pool selection in multi-agent systems
Timeline
- 2026First-author arXiv release of 'From Zero to Hero: An Open LLM Ecosystem for Armenian', reviewed as the first fully open, reproducible Armenian LLM ecosystem.
- 2026Middle-author arXiv paper 'Mo' Models, Mo' Problems', a systematic 23-model analysis of model-pool selection for LLM multi-agent systems.
- 2026Both papers entered the Evolve catalog with reproducibility assessments averaging 78 across 2 papers (high and medium readiness).
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.
- Profile is name-resolved from public bibliographic data; no verified institution, country, or ORCID is available.
- Bibliometric measures derive from OpenAlex-based public data; only the 2 Evolve-reviewed papers (both 2026 arXiv preprints) inform this profile.
- Output, impact, and activity dimensions lack underlying data (publication counts, citations, yearly cadence) and are intentionally left unscored.
- The reproducibility mean of 78 is computed over just 2 papers, so the score rests on a narrow base.