Papers, preprints, code, and talks from the OOAARG research group.https://ooaarg.org/en-ushttps://ooaarg.org/publications/oon-2026/https://ooaarg.org/publications/oon-2026/A second-order online learner that matches the regret of the best Newton step in hindsight, while remaining one-pass with O(d²) memory.Sat, 02 May 2026 00:00:00 GMTOralbanditsSecond-orderRegretConvexLeila Aydın, Maya Singh, Renat Ostrovskyhttps://ooaarg.org/publications/card-2026/https://ooaarg.org/publications/card-2026/Calibrated cardinality estimation under workload drift — confidence intervals that hold without distributional assumptions.Tue, 28 Apr 2026 00:00:00 GMTPaperdbmsCardinalityCalibrationDriftJihoon Park, Aylin Demir, Sasha Volkovhttps://ooaarg.org/publications/mirror-heavy/https://ooaarg.org/publications/mirror-heavy/A mirror descent variant whose potential adapts to the heaviness of the loss tail, achieving instance-optimal regret without prior tail knowledge.Tue, 14 Apr 2026 00:00:00 GMTPreprintbanditsMirror DescentHeavy tailsLeila Aydın, Maya Singh, Renat Ostrovskyhttps://ooaarg.org/publications/price-adapt/https://ooaarg.org/publications/price-adapt/Adaptive regret bounds come at a constant — but unavoidable — multiplicative cost. We pin down the exact frontier and exhibit matching algorithms.Thu, 02 Apr 2026 00:00:00 GMTJournalbanditsAdaptiveLower boundsMaya Singh, Renat Ostrovskyhttps://ooaarg.org/publications/oco-bench/https://ooaarg.org/publications/oco-bench/14 environments, 23 algorithms, one CLI. Reproducible regret curves with seed-level controls and standard plotting harnesses.Sun, 22 Mar 2026 00:00:00 GMTCodebanditsBenchmarkOpen-sourceJihoon Park, Leila Aydınhttps://ooaarg.org/publications/open-bandit/https://ooaarg.org/publications/open-bandit/How much harder is bandit learning when the action set is unbounded? Surprisingly, only by a logarithmic factor — provided the loss is well-conditioned.Thu, 04 Dec 2025 00:00:00 GMTSpotlightbanditsOpen vocabularySpotlightLeila Aydın, Jihoon Park, Sasha Volkov, Renat Ostrovskyhttps://ooaarg.org/publications/ada-buf/https://ooaarg.org/publications/ada-buf/A buffer manager that treats page replacement as a contextual bandit, with regret guarantees against the best fixed policy in hindsight.Tue, 18 Nov 2025 00:00:00 GMTPaperdbmsBuffer poolContextualAylin Demir, Jihoon Park, Sasha Volkovhttps://ooaarg.org/publications/log-regret/https://ooaarg.org/publications/log-regret/When both players play strongly convex losses, an extragradient method attains O(log T) duality gap with high probability.Sat, 12 Jul 2025 00:00:00 GMTPaperbanditsSaddleExtragradientMaya Singh, Renat Ostrovsky, Leila Aydınhttps://ooaarg.org/publications/talk-colt/https://ooaarg.org/publications/talk-colt/A 45-minute survey of recent progress on parameter-free and adaptive online learning, given as a plenary at COLT 2025.Sun, 15 Jun 2025 00:00:00 GMTTalkbanditsTalkSurveyRenat Ostrovskyhttps://ooaarg.org/publications/q-opt/https://ooaarg.org/publications/q-opt/A plan-selection bandit that achieves sublinear regret against the best plan per query template, even under heavy-tailed latency.Sun, 01 Jun 2025 00:00:00 GMTPaperdbmsQuery plansOnlineAylin Demir, Sasha Volkovhttps://ooaarg.org/publications/index-learn/https://ooaarg.org/publications/index-learn/A self-correcting learned index whose error stays bounded as the underlying key distribution drifts, without retraining from scratch.Tue, 22 Oct 2024 00:00:00 GMTPaperdbmsLearned indexesDriftJihoon Park, Aylin Demirhttps://ooaarg.org/publications/sgd-stable/https://ooaarg.org/publications/sgd-stable/A unified stability framework that recovers and tightens generalization bounds for SGD under online and i.i.d. settings.Fri, 03 May 2024 00:00:00 GMTJournalbanditsStabilityGeneralizationMaya Singh, Leila Aydın