Early-career Clemson University researcher in randomized numerical linear algebra for text mining and LLM evaluation
2
papers here
1
reviewed by Evolve
2
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: the indexed record is too thin for publication-volume measures. This axis would track total output, annual publication rate, and authorship roles across indexed venues.
Impact—
Not yet scored: citation-based measures need an accumulation window this early record lacks. The axis would measure citation counts, aggregate indices, and influence on downstream randomized-SVD and co-clustering work.
Focus100/100
Scores 100, and the raw metric explains it: the field distribution places 100% of the field-classified record in one field — machine learning / text mining with a randomized numerical linear algebra emphasis. One caveat: only one cataloged paper currently carries a field label, so the perfect concentration partly reflects a thin classification base.
Collaboration14/100
Scores 14, consistent with the raw co-authorship count: exactly 1 distinct co-author appears across the indexed record, with Carlos Toxtli the sole frequently-listed collaborator. Both cataloged papers list the researcher as first author and not corresponding, so the record reflects small, tightly-scoped two-person collaboration rather than extended team science.
Activity—
Not yet scored: the snapshot lacks the multi-year cadence data this axis needs. It would measure publication frequency, year-over-year consistency, and gaps between indexed releases such as the two 2026 arXiv preprints.
Reproducibility55/100
Scores 55, computed as the mean readiness across 1 assessed paper. The evaluated preprint, on randomized SVD approximations for spectral co-clustering, carries a medium readiness rating, and Evolve's verdict calls the empirical study honest but thin with significance testing absent. Because the score rests on one paper only, read it as a narrow-sample estimate rather than a stable track record.
Research summary
Fateme Mazdarani is a researcher at Clemson University whose cataloged work sits at the intersection of randomized numerical linear algebra and text mining, with a second line extending into evaluation of large language models. The Evolve catalog lists two 2026 arXiv preprints, both first-authored. The first, "Randomized SVD Approximations for Spectral Co-Clustering of Word-Document Matrices," compares two randomized SVD approximations — random projection versus element-wise matrix sampling — as accelerators for spectral co-clustering of word-document matrices, framed as bipartite graph partitioning. Evolve's verdict characterizes the empirical study as honest but thin, noting that significance testing is absent, and assigns a medium reproducibility-readiness rating; the paper is the sole basis for the researcher's reproducibility score of 55. The second preprint, "Beyond the Answer Key," evaluates the robustness of large language models for step-level mathematical verification, though it has not yet been reviewed and carries no verdict, keywords, or readiness rating. Bibliometrically the profile is early-stage: field classification assigns 100% of the labeled record to a single field, producing a focus score of 100; only 1 distinct co-author (Carlos Toxtli) appears, yielding a collaboration score of 14; and output, impact, and activity remain unscored for lack of sufficient data. The overall picture is a focused, early-career researcher whose trajectory is best read from a small number of recent, first-authored preprints.
Collaboration pattern
The indexed record shows a minimal collaboration footprint: 1 distinct co-author in total, with Carlos Toxtli the only recurring collaborator. Both cataloged papers are first-authored without a corresponding-author role, consistent with a small-team working pattern. As indexed output grows, the co-author count and any shift toward corresponding roles will be the clearest signals of an expanding collaborative network.
Research keywords
Highlights
Two randomized SVD routes for faster spectral co-clustering
Honest but thin empirics, medium reproducibility readiness
LLMs as step-level math verifiers under robustness evaluation
The second first-authored 2026 arXiv preprint, "Beyond the Answer Key," examines how robustly large language models verify intermediate solution steps rather than only final answers; it is cataloged but not yet reviewed, so no verdict, keywords, or readiness rating exist for it.
Timeline
- 2026First-authored arXiv preprint on randomized SVD approximations for spectral co-clustering of word-document matrices.
- 2026First-authored arXiv preprint on robustness evaluation of large language models for step-level mathematical verification.
- 2026Evolve reviews the co-clustering preprint, rating its reproducibility readiness as medium and noting the absence of significance testing.
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 OpenAlex-derived bibliometric data only and reflect a thin early-stage record.
- Output, impact, and activity dimensions are unscored because the underlying bibliometric data is insufficient.
- The second cataloged paper (LLM step-level verification) has not yet been reviewed by Evolve, so it carries no verdict, keywords, or reproducibility rating.
- No ORCID is provided; the profile is name-resolved from public bibliographic data, so identity disambiguation is limited.