Featured · ICML 2026 · Paper

Learned Cardinality with Distribution-Free Confidence

Calibrated cardinality estimation under workload drift — confidence intervals that hold without distributional assumptions.

02 / Subgroups

What we work on.

We organize around problems, not techniques. Each thread carries the same commitment to provable guarantees and practical deployment.

Bandits & Online Learning

Sequential decision-making under partial information — multi-armed bandits, contextual bandits, and adversarial online learning with provable regret guarantees.

UCBThompson SamplingMirror DescentAdversarialRegret bounds
Õ(√T)
tight regret
running average across our 2024–2026 contributions
03 / Latest publications

Out of the lab.

New preprints, conference papers, and code releases — most recent first. Each one comes with a one-click citation.

All publications →
04 / Join us

Let's work together.

We do science for both academia and industry, and we read every application. Whether you're a prospective student, a collaborator, or a company with a hard problem, we'd love to hear from you.