Vaneet Aggarwal
Purdue University West Lafayette · United States
Online learning theorist at Purdue developing tight regret bounds for constrained and submodular optimization problems.
4
papers here
2
reviewed by Evolve
2
first-author
1
corresponding
—
works · OpenAlex
3
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—
No output score is available: the OpenAlex record does not provide reliable aggregate publication totals for this name, so this axis — which would measure the volume, velocity, and consistency of publications over time — is left unscored.
Impact—
Impact is not scored: citation-based aggregates are too sparse in this record, where only h-index 3 and i10-index 1 are available. This axis would measure citation influence and uptake of the work by other researchers.
Focus0/100
Focus scores 0, the lowest concentration mark. The field distribution is split exactly evenly — 0.5 for online learning and optimization theory and 0.5 for machine learning theory (online optimization) — and topics fan out across five clusters: reinforcement learning in robotics (20), bandit algorithms (20), stochastic gradient optimization (14), optimization and search problems (9), and graph algorithms and complexity (7), with catalog papers ranging from pure regret theory to LLM-agent evaluation methodology.
Collaboration41/100
The collaboration dimension scores 41. The OpenAlex record lists 7 distinct co-authors, and Evolve's frequent-coauthor set comprises Yiyang Lu, Mohammad Pedramfar, Yang Xu, Chenang Li, Jiefu Zhang, Haixiang Sun, and Zhou Li. Across the four 2026 catalog papers, the researcher alternates between first-author theory papers and last-author positions on agent-evaluation and decentralized-optimization projects, a pattern consistent with a small, stable core group rather than a large rolling network.
Activity100/100
Activity scores a perfect 100. The record shows 35 works in the last three years, and the counts-by-year series jumps from 1 work in 2025 to 34 works dated 2026 — the dominant spike in the snapshot. All four Evolve catalog papers are 2026 arXiv postings, confirming the burst is current rather than historical, though such a sudden concentration also warrants identity verification.
Reproducibility55/100
Reproducibility scores 55, computed as the mean over 2 scored papers. Evolve's reviews describe both first-author theory papers as rigorous but theory-only — one lacks empirical validation, the other gives asymptotic, assumption-dependent guarantees — and rate all three reviewed papers at medium reproducibility readiness, so the score reflects solid mathematical specification paired with limited empirical grounding.
Research summary
Vaneet Aggarwal is a researcher at Purdue University (West Lafayette, US) whose recorded output centers on the theory of online learning and optimization. The profile is name-resolved from OpenAlex bibliographic data. The field distribution splits exactly evenly — 0.5 for online learning and optimization theory and 0.5 for machine learning theory (online optimization) — with topical mass in bandit algorithms (20), reinforcement learning in robotics (20), and stochastic gradient optimization (14). Four 2026 arXiv papers sit in the Evolve catalog. Two are first-author theory results: a constrained online learning analysis proving O(√T) regret and O(√(T log T)) budget violation under noisy constraints with matching lower bounds, and an online non-monotone DR-submodular maximization algorithm claiming a 0.401 approximation factor that breaks the 1/e barrier. Two are last-author projects: the SASST protocol for selection-aware stress testing of LLM agents, and Dec-BFTRL, a decentralized online optimization method with square-root regret. Recorded activity is extreme — 35 works in the last three years, 34 of them dated 2026 — while the citation footprint remains tiny (h-index 3, i10-index 1). Seven distinct co-authors recur, and reproducibility scoring over two papers averages 55, with Evolve noting rigorous theory but limited empirical validation.
Collaboration pattern
The OpenAlex record lists 7 distinct co-authors, and Evolve's frequent-coauthor set comprises Yiyang Lu, Mohammad Pedramfar, Yang Xu, Chenang Li, Jiefu Zhang, Haixiang Sun, and Zhou Li. Across the four 2026 catalog papers, the researcher alternates between first-author theory work and last-author positions on agent-evaluation and decentralized-optimization projects, a pattern consistent with a small, stable core group (collaboration score: 41) rather than a large rolling network.
Research keywords
Highlights
Constrained online learning under noisy constraint values
Online non-monotone DR-submodular maximization at the 0.401 factor
SASST: selection-aware stress testing for interactive agents
Dec-BFTRL: square-root regret for decentralized online optimization
Last-author 2026 arXiv paper on decentralized online upper-linearizable optimization under separation access, achieving square-root regret with an application to continuous submodular maximization; listed in the catalog without a full Evolve review yet.
Timeline
- 2022Earliest year in the bibliometric snapshot: OpenAlex records 3 works for this profile.
- 2023Recorded output rises modestly to 4 works in the OpenAlex snapshot.
- 2025The counts-by-year series records just 1 work for this year.
- 2026Output bursts to 34 dated works; four arXiv papers enter the Evolve catalog — two as first author (constrained online learning; DR-submodular maximization) and two as last author (SASST; Dec-BFTRL).
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 and raw counts come from a single OpenAlex record resolved by name (ORCID not linked), so the record may be incomplete or fragmented.
- Output and impact dimensions are unscored because aggregate publication and citation data are insufficient; h-index 3 and i10-index 1 are the only citation signals available.
- Evolve review verdicts, keywords, and slugs exist for only 3 of the 4 catalog papers; Dec-BFTRL is listed without a full review.
- The reproducibility mean (55) is based on only 2 scored papers and may not represent the broader body of work.