University of Pisa NLP researcher spanning LLM evaluation, mechanistic interpretability, and topic modeling.
2
papers here
1
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
7
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—
No output score is computed yet: Evolve has not consolidated a verified, de-duplicated publication list for this OpenAlex identity. This axis would measure total publication volume, venue quality, and productivity trends once the full record is assembled.
Impact—
No impact score is available yet: citation uptake for this researcher has not been verified against a consolidated record. This axis would measure citation counts, h-index trajectory, and influence on subsequent work once the data is sufficient.
Focus100/100
The focus score of 100 matches an exceptionally concentrated topical profile: OpenAlex topic counts put Topic Modeling first at 21 works, followed by Misinformation and Its Impacts (12), Spam and Phishing Detection (11), Natural Language Processing Techniques (10), and Sentiment Analysis and Opinion Mining (6) — adjacent text-analysis themes under one NLP umbrella, consistent with the single indexed field of natural language processing / mechanistic interpretability.
Collaboration32/100
The collaboration score of 32 reflects a deliberately small working circle: only 4 distinct coauthors appear in the recent OpenAlex window, and the recurring names — Lucia Passaro, Alessandro Lenci, Serena Auriemma, and Davide Bacciu — recur across the record. This is a tight, stable in-group pattern rather than a broad international co-authorship network, consistent with both 2026 first-author preprints.
Activity96/100
Activity is the standout axis, scored 96: OpenAlex counts 26 works over the recent three-year window, with annual output climbing from 3 works in 2020 to 6 in 2024, peaking at 15 works in 2025, and 5 already recorded in 2026. That decade-long trajectory from 1 work in 2015 to sustained double-digit output marks a clear acceleration.
Reproducibility100/100
The reproducibility score of 100 derives from Evolve's assessment of 1 catalog paper with a mean repro-readiness of 100 (reproPapers: 1). The SAE interpretability study is rated high reproducibility readiness and the likelihood-ranking study medium; both are first-author preprints that state their method limits explicitly, supporting the top score on a still-thin base.
Research summary
Alessandro Bondielli is a natural language processing researcher at the University of Pisa whose profile, name-resolved from public OpenAlex bibliographic data, connects classical text mining with frontier LLM science. His topic record — Topic Modeling (21 works), Misinformation and Its Impacts (12), Spam and Phishing Detection (11), Natural Language Processing Techniques (10), and Sentiment Analysis and Opinion Mining (6) — forms a text-analysis core that has recently pivoted toward how large language models are evaluated and how they represent language internally. The two Evolve-reviewed papers are both 2026 first-author arXiv preprints: a large study across 95 LLMs showing statement-likelihood accuracy stays flat with scale while prompted accuracy soars, and a mechanistic interpretability study locating part-of-speech structure in LLaMA-3 sparse-autoencoder latents. The OpenAlex snapshot reports an h-index of 7 and i10-index of 6, 26 works in the recent three-year window, and a sharp rise to 15 works in 2025 with 5 already recorded in 2026. Evolve scores focus and reproducibility at 100 and activity at 96, while output and impact scoring is still pending. Collaboration is compact: 4 distinct coauthors, including Lucia Passaro, Alessandro Lenci, Serena Auriemma, and Davide Bacciu. The overall picture is a methodologically careful researcher in a phase of clear acceleration.
Collaboration pattern
The collaboration score of 32 reflects a compact circle: only 4 distinct coauthors appear in the recent OpenAlex window, and the recurring names are Lucia Passaro, Alessandro Lenci, Serena Auriemma, and Davide Bacciu. Both Evolve-reviewed 2026 papers are first-author works without a corresponding-author role, suggesting close, stable in-group partnerships rather than a broad international co-authorship network.
Research keywords
Highlights
Likelihood ranking doesn't scale like prompting
Part-of-speech structure emerges in SAE latents
Output accelerates to a 2025 peak
OpenAlex counts 15 works in 2025 and 5 already in 2026, against 26 works in the recent three-year window — a sharp acceleration from 6 works in 2024.
Timeline
- 2015First work recorded in the OpenAlex window (1 work).
- 2019–2020Output grows slowly — 2 works in 2019 and 3 in 2020.
- 2021–2022A steadier cadence sets in, with 4 works in each of 2021 and 2022.
- 2024Six works recorded, the strongest pre-surge year in the record.
- 2025Peak productivity: 15 works recorded in a single year.
- 2026Five works already recorded, including both Evolve-reviewed first-author arXiv preprints on likelihood ranking and SAE part-of-speech structure.
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.
- Bibliometrics come from the OpenAlex snapshot only (identity: openalex); the profile is name-resolved from public bibliographic data, not a verified CV.
- Output and impact dimension scores are not yet computed by Evolve; the raw h-index (7) and i10-index (6) are shown but unscored.
- The reproducibility score of 100 rests on a single assessed paper (reproPapers: 1), so the evidence base is thin.
- Evolve catalog coverage is limited to two 2026 arXiv preprints; the researcher's other works have not been reviewed yet.