Evolve.Pub
Sign inScan your paper
CC

Claudiu Creanga

Herrick Technology Laboratories (United States)

Unclaimed · AI-structured profileOpenAlex

NLP researcher building LLM evaluation benchmarks — from question archaeology to biomedical relation extraction.

2

papers here

2

reviewed by Evolve

2

first-author

1

corresponding

—

works · OpenAlex

—

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-09-20) 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: insufficient dataOutputImpact: insufficient dataImpactFocus: 0/100FocusCollaboration: 32/100CollaborationActivity: insufficient dataActivityReproducibility: 38/100Reproducibility

Output—

Data not yet available. This axis would measure publication volume and productivity — total outputs, authorship roles and output rate over time. The catalog currently holds only two 2026 arXiv papers, too small a base to compute a score.

Impact—

Data not yet available. This axis would measure scholarly influence — citations, aggregate indices and uptake of the released benchmarks. Neither reviewed paper carries citation data, so no impact score is computed here.

Focus0/100

The focus score is 0, a direct consequence of the raw field distribution: activity splits exactly 50/50 between natural language processing (computational linguistics) and biomedical NLP/computational biology. The two reviewed 2026 papers form a deliberate dual track — one general LLM-evaluation benchmark and one biomedical relation-extraction benchmark — sharing an evaluation methodology rather than a single narrow specialisation.

Collaboration32/100

The collaboration score is 32, based on 4 distinct co-authors in the raw data. The pattern is compact and repeated: Liviu P. Dinu appears twice in the co-author records and Teodor Marchitan once, and both reviewed papers are first-authored — one also corresponding — suggesting Creanga leads small writing teams rather than joining large consortia. No large-scale or cross-institutional collaboration is visible in the current evidence.

Activity—

Data not yet available. This axis would measure publishing cadence and continuity — papers per year and recency of output. With only a single publication year (2026) in evidence, no trend or activity level can be established.

Reproducibility38/100

Reproducibility scores 38, the mean over two rated papers. Evolve rates the question-archaeology study at medium readiness — usable, but with under-specified headline metrics and statistics — and the biomedical relation-extraction benchmark at low, with SOTA margins reported without significance tests. Both papers are benchmark-driven, which typically aids replication, but reporting rigor is the current limiting factor.

Research summary

Claudiu Creanga is a researcher at Herrick Technology Laboratories (United States) whose work sits at the interface of general natural language processing and biomedical NLP; the bibliometric field distribution splits evenly, 0.5 to each. The evidence base is compact but coherent: two 2026 arXiv papers, both first-authored, both framed as evaluations of large language models on demanding tasks. 'Reading Between the Lines: Can LLMs Discover the Question Behind the Text?' introduces question archaeology — inferring a text's genesis question — on a 335-text benchmark where Gemini outperforms humans, though Evolve's review flags under-specified headline metrics and statistics. The second paper, first- and corresponding-authored, benchmarks LLMs, masked language models and a hybrid approach on the SNPPhenA corpus for biomedical relation extraction, finding that few-shot O1 and fine-tuned Gemini 2.0 Pro lead all three tasks, while SOTA margins lack significance tests. Collaboration is small-scale — four distinct co-authors, including Liviu P. Dinu and Teodor Marchitan, yielding a collaboration score of 32 — and reproducibility readiness is modest, with a mean score of 38 across two rated papers (medium and low). Output, impact and activity metrics are not yet available, making this an early-stage profile.

Collaboration pattern

Collaboration is small-team and repeated: the raw data lists 4 distinct co-authors, with Liviu P. Dinu appearing twice and Teodor Marchitan once. Both catalog papers are first-authored and one is also corresponding, indicating a lead-author role within compact groups rather than large consortia — consistent with the low-moderate collaboration score of 32.

Research keywords

question archaeologyauthorial intentLLM evaluationbenchmark datasetquestion generationreverse promptingbiomedical relation extractionSNP-phenotype associationlarge language modelsSNPPhenA corpusmasked language modelsfew-shot learning

Highlights

Timeline

  1. 2026Published 'Reading Between the Lines: Can LLMs Discover the Question Behind the Text?' on arXiv as first author, introducing question archaeology on a 335-text benchmark.
  2. 2026Published the biomedical relation-extraction benchmark on the SNPPhenA corpus as first and corresponding author, comparing LLMs, masked language models and a hybrid approach.
  3. 2026Established a dual-track LLM evaluation agenda spanning general NLP and biomedical NLP, reflected in an even 50/50 field distribution.

Papers in the Evolve catalog

Frequent co-authors

Liviu P. Dinu×2Claudiu Creanga×1Liviu P. Dinu×1Teodor Marchitan×1

Is this you? Claim this profile

Enter the email you published with. We'll send a one-time claim link; after claiming you can edit the introduction or hide this page. Requests from a mismatched address go to manual review.

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 indicators are computed from OpenAlex-derived data only; no citation counts, h-index or venue-level metrics were provided in the input.
  • Profile is name-resolved from public bibliographic data; no ORCID or country field was supplied, so identity confidence is limited.
  • Output, impact and activity dimensions lack underlying data and are deliberately unscored; only two reviewed papers inform this profile.
  • The frequent-coauthor list contains duplicate entries and the profile's own name; these are treated as data artifacts and excluded from the analysis.
Claudiu Creanga — Researcher Profile | Evolve.Pub