Tag
This paper studies KL-regularized contextual bandits and shows that greedy sampling can achieve logarithmic regret without explicit eluder-dimension dependence for both reward and preference feedback.
This paper proves that Thompson sampling achieves polynomial regret for non-monotone convex ridge bandits, showing that monotonicity is not necessary, with a new regret bound of Õ(d^{9/2} √n).
This paper proposes new reoptimization algorithms for contextual bandits with knapsack constraints, achieving an average regret bound of O((ln T)^3 / T) and improving existing results.
This paper studies multi-agent multi-armed bandits with heavy-tailed rewards under three information-asymmetry regimes, proposing robust decentralized algorithms with regret guarantees nearly matching centralized rates, and validating them on Pareto-distributed reward environments.
This paper introduces a discrepancy-rounding framework for stochastic bandits with exact minimum-exposure constraints, achieving fair regret governed by the nonmandatory budget rather than horizon. It proposes algorithms with minimax and instance-dependent optimality guarantees, handles time-varying and overlapping group floors, and validates through experiments.
This paper studies stochastic linear bandits where the agent only observes a random subset of action coordinates, proving that sublinear regret is possible when actions have low intrinsic dimension, and proposes the TOFU-POV algorithm with theoretical guarantees.
This paper demonstrates that the regret of Bayesian and multiplicative-weights updates satisfies an exact information-accounting identity, decomposing the learner's excess loss into an uncertainty payment and a reduction in information distance to any comparator. The cumulative payment defines intrinsic time, leading to exact adaptive regret decompositions that unify Hedge, Bayesian model averaging, online convex optimization, and other algorithms.
This paper formalizes embedding model routing as an adversarial contextual linear bandit with low-rank experts, proposing the Hypentropy Policy Gradient (HPG) algorithm that achieves O~(s√(MT)) policy regret, avoiding the curse of dimensionality.
This paper formalizes Streaming Knowledge Compilation for LLM wikis, introducing a materiality signal to proactively pin important documents from a streaming corpus under a token budget. It proves an O(√(T log K)) regret bound and validates the approach in finance and Wikipedia domains, showing that regret analysis is a reliable evaluation metric.
This paper introduces an online contextual Pandora's Box model for adaptively querying and selecting LLM APIs, proposing a learning approach that combines GMM estimation with UCB-style confidence bounds and proving dimension-dependent regret bounds.
Proposes a truthful online preference aggregation mechanism for LLM fine-tuning in mobile crowdsourcing, addressing strategic worker misreporting and achieving sublinear regret.
This note presents a research moment where Codex helped find a new rare-switching rule for private linear bandits, using the generalized Rayleigh quotient to overcome the failure of determinant-based monotonicity due to Gaussian noise.