Roy Zavida

—

Unclaimed · AI-structured profileOpenAlex

Early-career researcher building deterministic synthetic enterprise benchmarks for evaluating LLM agents

papers here

reviewed by Evolve

first-author

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—

The output axis (publication volume and trajectory over time) is not yet scored: the catalog captures only three works, all posted to arXiv in 2026, so there is no multi-year volume record from which to compute a stable output trend.

Impact—

Impact is not yet scored: the recorded h-index and i10-index are both 0, consistent with papers published in 2026 that have had little time to accrue citations; this axis would measure citations received and downstream scholarly influence.

Focus0/100

The focus score is 0 because the field distribution splits exactly into thirds: LLM agent benchmarking (AI/NLP), AI — LLM agent evaluation, and data systems/synthetic data generation for agentic AI each take 0.333. In substance these are one program around the Era by Eon estate; OpenAlex topics put Topic Modeling first (3 works), then AI applications (2), and NLP techniques, privacy-preserving technologies, and AI planning (1 each).

Collaboration39/100

Collaboration scores 39, based on 6 distinct coauthors across the three 2026 papers. Roy Zavida holds a middle author position on every paper with no corresponding-author role, indicating a contributor within a small, stable team rather than a solo or leading effort; recurring partners include Benjamin Gruenbaum, Doron Porat, Assaf Natanzon, Chen Dinachi, Or Itzahary, and Omer Niv.

Activity40/100

Activity scores 40: the counts-by-year record shows all 3 works appearing in 2026, and recent3yWorks is 3, so publishing is concentrated in a single year with no earlier record in this snapshot — an early-stage publishing rhythm that does not yet support a multi-year cadence assessment.

Reproducibility43/100

Reproducibility scores 43 as the mean readiness rating across 3 reviewed papers (two rated medium, one low). Evolve's verdicts cite concrete gaps: a single-seed evaluation and internal count errors in the hidden-knowledge paper, circular evaluation without external validation in the synthesis paper, and one small estate with unreleased code in the benchmark paper.

Research summary

Roy Zavida is an early-stage researcher working at the intersection of LLM agent evaluation and synthetic data engineering; this profile is built from OpenAlex bibliographic records and three 2026 arXiv preprints in the Evolve catalog. The three papers form one coherent program around the 'Era by Eon' benchmark family: 'The Era by Eon Benchmark' introduces seed-generated fictional enterprise estates with exact ground-truth answer keys for benchmarking LLM agents; 'Generating a Consistent Enterprise' builds deterministic multi-system business data with reference-free evaluation; and 'Era by Eon: Benchmarking Enterprise Agents on Hidden Knowledge' adds eight deterministic hidden-knowledge questions to probe agents beyond surface retrieval. Zavida holds a middle author position on all three papers without corresponding-author status, collaborating with a compact group of six named co-authors. Evolve's reviews credit the construct design as elegant and the engineering as sound, while flagging recurring evaluation weaknesses: single-seed runs, internal count errors, circular reference-free assessment, and unreleased code. The recorded h-index and i10-index are both 0, consistent with a 2026-only publication record (3 works, 3 works in the last three years), and reproducibility readiness averages 43/100 across the three papers. This is an emerging, technically specific profile whose citation impact is not yet measurable.

Collaboration pattern

All three 2026 papers are co-authored with the same compact team: Benjamin Gruenbaum, Doron Porat, Assaf Natanzon, Chen Dinachi, Or Itzahary, and Omer Niv — 6 distinct coauthors in total. Roy Zavida holds a middle author position on every paper with no corresponding-author role, consistent with a specialized contributor inside a stable research group rather than an independent lab lead.

Research keywords

LLM agent benchmarkingLLM agent evaluationsynthetic enterprise datadeterministic data generationexact ground truthhidden knowledge question answeringreference-free evaluationtool use evaluationquestion-conditioned generationsystems of recordbenchmark constructionreproducible evaluation

Highlights

Seed-generated enterprise estate with exact ground truth
The Era by Eon benchmark generates fictional companies from seeds so that every agent question has an exact answer key; Evolve rates the validation as sound but notes that the headline evaluation covers one small estate and the code is unreleased.
Eight hidden-knowledge questions for enterprise agents
This work extends Era by Eon with 8 deterministic hidden-knowledge questions testing agents beyond surface retrieval; Evolve calls the construct elegant and the stated limits honest, but flags a single-seed evaluation and internal count errors.
Deterministic multi-system business data with reference-free evaluation
Generates a consistent synthetic enterprise spanning multiple systems of record and evaluates it without reference data; Evolve finds the engineering sound but the evaluation circular and lacking external validation.

Timeline

2026First cataloged work: 'The Era by Eon Benchmark' introduces a generated enterprise estate with exact ground truth for benchmarking LLM agents.
2026Follow-up work: 'Generating a Consistent Enterprise' details deterministic synthesis and reference-free evaluation of multi-system business data.
2026Third work: 'Era by Eon: Benchmarking Enterprise Agents on Hidden Knowledge' adds 8 deterministic hidden-knowledge questions to the benchmark.

Papers in the Evolve catalog

Frequent co-authors

Benjamin Gruenbaum×3 Doron Porat×3 Assaf Natanzon×3 Chen Dinachi×3 Or Itzahary×3 Omer Niv×1

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.

Identity is name-resolved from public OpenAlex bibliographic data; institution, country, and ORCID are unverified in this profile.
All bibliometrics come from a single OpenAlex snapshot; the h-index (0) and i10-index (0) reflect a 2026-only publication record and are not yet meaningful.
Qualitative assessments quote Evolve editorial reviews of three arXiv preprints; they are not peer-review verdicts and may change as the works are revised.
The publication record (3 works, all 2026) is too small for trend analysis, which is why the output and impact axes are left unscored.