Clemson University researcher working on randomized numerical linear algebra and LLM verification robustness.
2
papers here
1
reviewed by Evolve
0
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 underlying publication-volume data — counts of papers, venues, and authorship roles over time — is not available in the current snapshot, so no output estimate is offered.
Impact—
Not yet scored: citation counts, citation trajectories, and downstream influence metrics are absent from the snapshot, so impact cannot be assessed; this axis would measure how widely the work is used and cited.
Focus100/100
Focus scores a perfect 100: the raw field distribution lists exactly one indexed field — machine learning / text mining (randomized numerical linear algebra). The reviewed paper on randomized SVD for spectral co-clustering sits squarely in it, and the second paper's topic, robustness evaluation of LLMs for step-level mathematical verification, stays close to the same algorithmic-and-evaluation agenda rather than scattering across unrelated areas.
Collaboration14/100
Collaboration scores 14, derived from exactly one distinct coauthor in the raw data; Fateme Mazdarani is the sole frequent collaborator. Toxtli holds the last author position on both catalog papers without being corresponding author, consistent with a senior or supervisory role inside a small, essentially two-person working pattern rather than a large lab network.
Activity—
Not yet scored: the snapshot lacks per-year publication timestamps and ongoing submission records, so recent research activity cannot be quantified; this axis would measure publishing cadence and current engagement.
Reproducibility55/100
Reproducibility scores 55, computed as a mean (reproMean 55) over only one assessed paper (reproPapers 1), which carries medium repro readiness. Evolve's verdict calls the empirical study honest but thin, with significance testing absent — code and experimental setup appear followable, but statistical rigor remains the main gap to stronger reproducibility.
Research summary
Carlos Toxtli is a researcher at Clemson University whose Evolve catalog profile centers on randomized numerical linear algebra and its applications to text mining. The catalog includes two 2026 arXiv papers on which Toxtli appears in the last author position, without being marked corresponding author. The reviewed paper, "Randomized SVD Approximations for Spectral Co-Clustering of Word-Document Matrices," compares two randomized SVD approximations — random projection versus element-wise sampling — for accelerating spectral co-clustering of word-document matrices, framed as bipartite graph partitioning. Evolve's verdict describes the empirical study as honest but thin, noting that significance testing is absent, and assigns medium reproducibility readiness. The second 2026 paper, "Beyond the Answer Key," evaluates large language models for step-level mathematical verification from a robustness perspective and has not yet been reviewed by Evolve. Bibliometrically the profile is narrow but coherent: a focus score of 100 rests on a field distribution containing a single field — machine learning / text mining (randomized numerical linear algebra); a collaboration score of 14 rests on exactly one distinct coauthor, Fateme Mazdarani; and a reproducibility score of 55 is the mean over one assessed paper. Output, impact, and activity lack underlying data and are deliberately unscored.
Collaboration pattern
The raw bibliometric data lists a single distinct coauthor, and the only frequent collaborator is Fateme Mazdarani, giving a small, stable two-person working pattern. Toxtli holds the last author position on both 2026 arXiv papers without being corresponding author, consistent with a senior contributor role. No multi-institution or large-lab co-authorship network is visible in the current snapshot, so broader collaboration behavior cannot yet be characterized.
Research keywords
Highlights
Randomized SVD approximations for faster spectral co-clustering
Robustness evaluation of LLMs as step-level math verifiers
Second 2026 arXiv paper, "Beyond the Answer Key: Robustness Evaluation of Large Language Models for Step-Level Mathematical Verification," with Toxtli as last author. It examines how robustly large language models judge individual solution steps; Evolve has not yet published a verdict, keywords, or repro-readiness rating for it.
Timeline
- 2026Posts "Randomized SVD Approximations for Spectral Co-Clustering of Word-Document Matrices" to arXiv, appearing as last author.
- 2026Posts "Beyond the Answer Key: Robustness Evaluation of Large Language Models for Step-Level Mathematical Verification" to arXiv, again as last author.
- 2026Evolve reviews the co-clustering paper, rating its reproducibility readiness as medium and recording a reproducibility mean of 55 for the profile.
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.
- Dimension scores come from a single bibliometric snapshot; the output, impact, and activity axes lack underlying data and are left unscored.
- Only two papers (both 2026 arXiv preprints) are in the Evolve catalog; peer-reviewed venues, citation counts, and h-index were not provided.
- The reproducibility score of 55 is a mean over just one assessed paper and may not represent overall practice.
- Profile is name-resolved from public bibliographic data (ORCID 0000-0002-9430-2661); country and other biographical fields were absent from the input.