University of Bucharest NLP researcher with recent last-author work on LLM benchmarking and biomedical text mining
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
17
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-28) 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—
Publication volume metrics are not available in this snapshot, so no output score can be reported. This axis would normally measure total paper count, venue distribution, and annual productivity across the researcher's full career.
Impact—
No citation counts, h-index, or influence metrics are available yet, so impact cannot be scored. This axis would measure how strongly the researcher's publications are cited and referenced across the field over time.
Focus0/100
The recorded focus score is 0, and the underlying field distribution splits evenly (0.5/0.5) between natural language processing and biomedical NLP/computational biology. Both catalog papers fall in these two adjacent areas, one on LLM evaluation benchmarks and one on biomedical relation extraction, so the current picture reflects a two-field profile; the two-paper catalog makes this reading provisional.
Collaboration41/100
The collaboration score is 32, based on 4 distinct co-authors in the bibliometric snapshot. The frequent co-author list clusters around Claudiu Creanga and Teodor Marchitan, and both 2026 catalog papers list the researcher in the last author position without corresponding-author status. The pattern points to small, stable teams rather than large consortia.
Activity100/100
Recent publication cadence data is insufficient in this snapshot, so no activity score is reported. This axis would measure publishing rhythm, such as papers per year and gaps between releases, across the researcher's full indexed record.
Reproducibility32/100
The reproducibility score is 38, averaged over 2 papers assessed by Evolve. The question-archaeology benchmark paper carries medium reproducibility readiness, while the biomedical relation-extraction benchmark is rated low, with Evolve noting under-specified headline metrics and missing significance tests. The reported 335-text benchmark and the SNPPhenA corpus define the current reproducible base.
Research summary
Liviu P. Dinu is a researcher at the University of Bucharest whose Evolve profile is anchored by two 2026 arXiv papers, both with him in the last author position. The catalog splits evenly across two adjacent fields — natural language processing (computational linguistics) and biomedical NLP/computational biology — each at 0.5 in the field distribution. The first paper, 'Reading Between the Lines,' introduces 'question archaeology,' testing whether large language models can recover the question behind a text on a 335-text benchmark, with Gemini reportedly outperforming humans, though Evolve flags under-specified metrics. The second benchmarks large language models, masked language models, and a hybrid on the SNPPhenA corpus for biomedical relation extraction, finding O1 few-shot and a fine-tuned Gemini 2.0 Pro lead all three tasks, again without significance testing. Collaboration metrics show 4 distinct co-authors and a collaboration score of 32, with Claudiu Creanga and Teodor Marchitan recurring; the pattern suggests compact, stable teams. Reproducibility scores 38 across 2 assessed papers, rated medium and low, held back by missing statistical rigor. Output, impact, and activity axes lack sufficient data and are left unscored. This is a senior-position profile whose recent direction is clearly LLM evaluation, in both general and biomedical settings.
Collaboration pattern
Collaboration runs through a small, stable circle: the snapshot records 4 distinct co-authors and a collaboration score of 32, with Claudiu Creanga and Teodor Marchitan as recurring names on the 2026 papers. Both catalog entries place Dinu in the last author position without corresponding-author status, consistent with senior roles on compact teams that publish preprints on arXiv. No large-consortium or cross-institution patterns are visible in the available data.
Research keywords
Highlights
Question archaeology: LLMs infer the question behind a text
Benchmarking LLMs on biomedical relation extraction
Timeline
- 2026Last-author arXiv paper 'Reading Between the Lines: Can LLMs Discover the Question Behind the Text?' introduces question archaeology on a 335-text benchmark.
- 2026Last-author arXiv paper benchmarks LLMs, masked language models, and a hybrid for biomedical relation extraction on SNPPhenA; O1 few-shot and fine-tuned Gemini 2.0 Pro lead all three tasks.
- 2026Both catalog entries land in the same year, splitting evenly between general NLP and biomedical NLP, with reproducibility readiness rated medium and low.
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 come from an OpenAlex-derived snapshot; output, impact, and activity axes lack sufficient data.
- The profile rests on only 2 Evolve-reviewed papers (both 2026 arXiv preprints); earlier career output is out of scope.
- The frequent co-author list includes the researcher's own name, likely a de-duplication artifact in the source data.
- Identity is name-resolved from public bibliographic data (ORCID 0000-0002-7559-6756); country-level affiliation details are not provided.