MIT theorist of online learning, algorithmic game theory and performative prediction; rigorous first-author theory.
3
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) 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—
Output is unscored: this snapshot contains no publication-count or productivity metrics beyond the three 2026 arXiv preprints in the Evolve catalog. This axis would measure aggregate research output, such as total publications and yearly production rates.
Impact—
Impact is unscored: the snapshot provides no citation counts, h-index, or other influence metrics. This axis would measure how widely the researcher's papers are cited and adopted by follow-up work.
Focus0/100
The focus score is 0, reflecting an even 50/50 field split between online learning and algorithmic game theory, and machine learning theory (performative prediction). All three 2026 catalog papers are pure theory; the two threads — regret and equilibrium analysis versus performative feedback loops — share algorithmic-game-theoretic tools, so the diffuse split reflects breadth within one theoretical agenda.
Collaboration36/100
The collaboration score is 36, based on 5 distinct coauthors in the snapshot. The frequent-coauthor list — Ashkan Soleymani, Patrick Jaillet, Mingyang Liu, Asuman Ozdaglar, and Juan Carlos Perdomo — suggests a small, stable, theory-centered circle rather than broad cross-institutional networks; two of the three catalog papers list the researcher in a middle author position.
Activity—
Activity is unscored: this snapshot has no publication-timing or per-year productivity data beyond the three 2026 preprints. This axis would measure publishing cadence and sustained research activity over time.
Reproducibility55/100
Reproducibility readiness scores a mean of 55 across the 2 assessed papers, both rated medium. The catalog papers are rigorous theory with complete proofs but explicitly no experiments — including the Õ(d log(1/ε)) deployment bound with matching lower bound and the constant individual-regret result — so reproducibility hinges on proof verifiability rather than empirical replication.
Research summary
Gabriele Farina is a researcher at the Massachusetts Institute of Technology working in theoretical machine learning, at the intersection of online learning, algorithmic game theory, and performative prediction. The Evolve catalog holds three 2026 arXiv preprints by this researcher, all rigorous theory papers with no experimental components. As first author, they claim a Õ(d log(1/ε))-deployment algorithm for performatively stable mixtures with a matching lower bound, drawing on expected variational inequalities, cutting-plane methods, and minimax theorems. A second preprint shows that gradient descent, mirror descent, and follow-the-regularized-leader remain no-regret against exact-form deviations, while circulation phenomena impose a linear-regret obstruction. A third claims the first horizon-independent individual regret guarantees in N-player games via ECHO-OFTRL; Evolve notes the proofs are complete but the constants and learning rate are astronomical. The bibliometric snapshot splits the field distribution evenly (50/50) between online learning and algorithmic game theory, and machine learning theory (performative prediction), producing a focus score of 0. Collaboration is concentrated in a small circle of 5 distinct coauthors, and reproducibility readiness averages 55 across the 2 assessed papers. Citation, output, and activity metrics are not yet available.
Collaboration pattern
The snapshot records 5 distinct coauthors and a collaboration score of 36, pointing to a compact, theory-oriented circle. Recurring collaborators include Ashkan Soleymani, Patrick Jaillet, Mingyang Liu, Asuman Ozdaglar, and Juan Carlos Perdomo. Authorship is mixed: first author on the performative prediction paper and middle positions on the two online-learning/game-theory preprints, consistent with both leading and contributing roles within small teams.
Research keywords
Highlights
Minimal model deployments stabilize performative feedback loops
Exact-form regret for OGD, mirror descent and FTRL
First horizon-independent individual regret in general games
Timeline
- 2026First-author arXiv preprint 'Stabilizing Performative Feedback Loops with Minimal Model Deployments' claims a Õ(d log(1/ε))-deployment algorithm for performatively stable mixtures with a matching lower bound.
- 2026arXiv preprint 'Exact-Form Regret for Gradient Descent, Mirror Descent and Follow-the-Regularized-Leader' establishes no-regret guarantees against exact-form deviations, with a circulation-based linear-regret obstruction.
- 2026arXiv preprint 'Constant Individual Regret in General Games' claims the first horizon-independent individual regret in N-player games via ECHO-OFTRL, with complete proofs though astronomical constants.
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 derived from OpenAlex-sourced data only; no citation counts, h-index, or venue-level metrics are included in this snapshot.
- Output, impact, and activity axes are unscored because the underlying bibliometric data is insufficient; only collaboration, focus, and reproducibility could be computed.
- Reproducibility readiness rests on only 2 assessed papers (mean 55); all three catalog entries are 2026 arXiv preprints with no experiments, so empirical-reproducibility evidence is thin.
- The profile is name-resolved from public bibliographic data linked to ORCID 0000-0002-3976-0061; the Evolve catalog covers just three preprints, all from 2026.