NLP researcher; first author of DANTINOX, a unified JAX/Flax framework for multi-paradigm language modeling
1
papers here
1
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-21) 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 data is available yet. This axis would measure publication volume — counts of papers per year and venue types — to gauge productivity. With the catalog currently limited to one reviewed paper, no reliable productivity score can be computed.
Impact—
Impact data is insufficient at present. This axis would track citations, usage, and uptake of the work by other researchers to estimate scholarly influence. Until DANTINOX accumulates citation history, no meaningful impact score can be reported.
Focus100/100
The focus score is 100/100: every paper in the reviewed catalog — currently one work — falls in a single field, natural language processing (machine learning systems). This is total but thin concentration: DANTINOX unifies autoregressive decoding, masked discrete diffusion, and flow matching on one Transformer backbone, all within the same modeling sub-area. The score reflects breadth of zero rather than deliberate exclusivity, given only one indexed paper.
Collaboration28/100
Collaboration scores 28/100, with 3 distinct coauthors observed. All cataloged work is co-authored with a stable trio — Aleksandar Fontana, Giulio Rossolini, and Andrea Saracino — suggesting a small, tightly-knit team rather than an extended network. On DANTINOX, Marco Simoni is first author (non-corresponding). A wider collaboration footprint, cross-institution ties, and solo-authorship patterns cannot yet be assessed from one indexed paper.
Activity—
Activity cannot be scored with current data. This axis would measure recent publishing tempo — papers in the last few years and cadence between releases — to show whether research is ongoing. More indexed papers are needed before a score is meaningful.
Reproducibility100/100
Reproducibility is a perfect 100/100 (reproMean 100 across 1 paper). DANTINOX carries a high reproducibility rating: a JAX/Flax framework paper whose engineering is solid and claims honestly scoped, the kind of artifact-oriented work others can rebuild on. The perfect score rests on a single paper, so it should be read as a strong start rather than a sustained track record.
Research summary
Marco Simoni is an early-stage natural language processing researcher whose reviewed catalog to date centers on machine learning systems for language modeling. The evidence base is deliberately small: one reviewed paper, DANTINOX (arXiv, 2026), on which Simoni is first author. DANTINOX presents a unified JAX/Flax framework that runs autoregressive decoding, masked discrete diffusion, and flow matching on a single Transformer backbone — a systems-and-frameworks contribution putting three generative modeling paradigms under one comparable implementation. Evolve's verdict calls it a credible framework paper with solid engineering and honestly scoped claims, while cautioning that its small-scale, undertrained comparisons are suggestive rather than definitive. Its reproducibility readiness is rated high, and the reproducibility dimension scores a perfect 100 (reproMean 100 over 1 paper). Focus likewise scores 100, reflecting a catalog entirely contained in one field — natural language processing (machine learning systems) — though that concentration rests on a single work. Collaboration scores 28, with 3 distinct coauthors: a stable trio of Aleksandar Fontana, Giulio Rossolini, and Andrea Saracino. Output, impact, and activity dimensions are currently missing because the underlying bibliometric data — publication counts, citation history, and recent publishing tempo — is not yet sufficient. No institution, country, or ORCID is attached to this profile; it is name-resolved from public bibliographic data.
Collaboration pattern
The observable collaboration pattern is a compact, recurring trio: all cataloged work is co-authored with Aleksandar Fontana, Giulio Rossolini, and Andrea Saracino, giving 3 distinct coauthors and a collaboration score of 28. On DANTINOX Simoni is first author but not corresponding author. With only one indexed paper, it is impossible to tell whether this trio is a fixed research group or one of several circles; no cross-institution or international ties are visible in the current data.
Research keywords
Highlights
One Transformer backbone, three generative paradigms
Solid engineering, honestly scoped claims
High reproducibility readiness from day one
Timeline
- 2026Released DANTINOX on arXiv: a unified JAX/Flax framework for multi-paradigm language modeling
- 2026Served as first author (non-corresponding) on DANTINOX, the single reviewed paper in this catalog
- 2026Work entered the catalog in natural language processing (machine learning systems) with high reproducibility readiness
- 2026Co-authored with Aleksandar Fontana, Giulio Rossolini, and Andrea Saracino — the profile's 3 distinct coauthors
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 data and Evolve's reviewed-paper catalog, which currently contains only one paper (DANTINOX, 2026).
- No institution, country, or ORCID is linked; Marco Simoni is name-resolved from public bibliographic data and may be ambiguous as a name.
- Output, impact, and activity dimensions lack sufficient underlying data and are shown as missing rather than scored.
- The perfect 100 focus and reproducibility scores rest on a single reviewed paper and should be read as preliminary.