Early-career Stanford researcher bridging statistical physics of neural networks and language-model pretraining
2
papers here
2
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
3
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) 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: publication-volume data for this profile are insufficient. This axis would measure total and per-year output; the raw record currently shows only an h-index of 3 and small yearly counts (1–6 works per year, 2019–2026).
Impact—
Not yet scored: citation data are too sparse for a reliable impact estimate; the record lists h-index 3 and i10-index 3, consistent with early, minimally cited work. This axis would measure citations, field uptake, and scholarly influence.
Focus0/100
Scores 0/100 because the record shows no dominant field: the field distribution is an even 0.5/0.5 split between statistical physics of neural networks (machine learning theory) and language model pretraining. The OpenAlex topic list adds condensed-matter themes (graphene and applications, topological materials, superconductivity and magnetism, quantum many-body, each with 4–6 topic hits), further spreading topical mass. Best read as a deliberate two-track profile rather than unfocused output.
Collaboration41/100
Scores 41/100. The record lists only 7 distinct coauthors, and both Evolve-catalog papers are first-author works without corresponding-author status, pointing to small-team authorship. Recurring partners span both research tracks: Adithya Sriram, Kfir Dolev, Michael Y. Li, G. Bruno De Luca, Nourya Cohen, Noah D. Goodman, and Yoav Levine. This is a compact, stable circle rather than a large distributed collaboration network.
Activity64/100
Scores 64/100 on sustained but uneven output. The per-year series shows 1 work in 2019 and 2020, a peak of 6 in 2021, then 1 in 2022 and 2023, recovering to 2 in 2024, 4 in 2025, and 2 in 2026. The last three years contribute 8 works, the strongest sustained stretch since the 2021 peak.
Reproducibility55/100
Scores 55/100, the mean across 2 papers assessed by Evolve, both rated medium reproducibility readiness. The verdicts sit in that middle band: the self-play pretraining paper is called a small-scale but well-controlled proof of concept, while the Hopfield paper is judged theoretically rigorous but with Fashion-MNIST evidence that remains qualitative rather than quantitative.
Research summary
Aditya Cowsik is an early-career researcher at Stanford University whose OpenAlex record splits evenly (0.5/0.5) between two fields: statistical physics of neural networks (machine learning theory) and machine learning for language model pretraining. The Evolve catalog contains two 2026 first-author arXiv preprints. The first, Self-Play Pretraining with Zero Data, shows that self-play over a universal Turing machine program space yields predictable zero-shot scaling on natural data — a small-scale but well-controlled proof of concept touching on Solomonoff induction and universal prediction. The second, Hierarchical Prototype Emergence in Modern Hopfield Models, argues that prototypes arise as stable minima in dense associative memories holding hierarchically correlated patterns. The bibliometric footprint is early-stage: h-index 3, i10-index 3, and a per-year output series of 1, 1, 6, 1, 1, 2, 4, 2 works from 2019 to 2026, with 8 works in the last three years (activity score 64/100). Collaboration is compact — 7 distinct coauthors, including Adithya Sriram, Kfir Dolev, Michael Y. Li, G. Bruno De Luca, Nourya Cohen, Noah D. Goodman, and Yoav Levine (score 41/100). Focus scores 0 because output splits evenly across the two tracks. Reproducibility is assessed at 55/100 from two papers, both rated medium readiness.
Collaboration pattern
This researcher works in a compact circle: 7 distinct coauthors appear in the OpenAlex record (collaboration score 41/100), and both catalog papers are first-author works without corresponding status. Recurring partners include Adithya Sriram, Kfir Dolev, Michael Y. Li, G. Bruno De Luca, Nourya Cohen, Noah D. Goodman, and Yoav Levine, spanning the machine-learning-theory and pretraining tracks.
Research keywords
Highlights
Zero-data self-play pretraining with predictable scaling
Prototypes emerge as stable minima in modern Hopfield models
Timeline
- 2019First works appear in the OpenAlex record — 1 publication.
- 2021Most prolific year on record, with 6 publications.
- 2025Output rebounds to 4 publications, continuing the recovery that began in 2024 (2 works).
- 2026Two first-author arXiv preprints enter the Evolve catalog: Self-Play Pretraining with Zero Data and Hierarchical Prototype Emergence in Modern Hopfield Models.
Papers in the Evolve catalog
- first author
Self-Play Pretraining with Zero Data
- first author
Hierarchical Prototype Emergence in Modern Hopfield Models
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 a single OpenAlex record (identity: openalex); citation coverage may lag arXiv-only output.
- Output and impact dimensions are unscored; h-index 3 and i10-index 3 rest on a small early-career citation base.
- The OpenAlex topic profile also lists condensed-matter themes (graphene, topological materials, superconductivity); this profile is name-resolved from public bibliographic data and may aggregate namesakes.
- Reproducibility (55/100) is averaged over only 2 Evolve-assessed papers, both arXiv preprints.