Finite-Time Regret Analysis of Retry-Aware Bandits

• URL: http://arxiv.org/abs/2605.20854v1
• Date: Wed, 20 May 2026 07:44:43 GMT
• Title: Finite-Time Regret Analysis of Retry-Aware Bandits
• Abstract summary: We study a bandit algorithm motivated by retry-aware objectives that value the best outcome among multiple attempts.<n>Given a posterior over arm values, ReMax chooses a sampling distribution that maximizes the posterior expected maximum reward over $M$ virtual draws.
• Score: 8.812244373657764
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We study a stochastic bandit algorithm motivated by retry-aware objectives that value the best outcome among multiple attempts, such as pass@$k$ and max@$k$. Given a posterior over arm values, ReMax chooses a sampling distribution that maximizes the posterior expected maximum reward over $M$ virtual draws. Although this objective was introduced in reinforcement learning as an exploration mechanism under uncertainty, its regret properties in bandit problems have remained unclear. For Gaussian rewards and the first nontrivial case $M=2$, we characterize the optimal ReMax distribution through an expected-improvement balance condition and prove the first sublinear regret bound for ReMax. Our analysis separates the usual saturation behavior of suboptimal arms from a ReMax-specific underestimation effect, in which the optimal arm may be sampled too rarely after an unfavorable estimate. This explains why ReMax can be more exploitative than Thompson sampling (TS) and why its regret analysis is technically delicate. Experiments support this picture: ReMax often outperforms KL-UCB and Thompson sampling under mild underestimation, while posterior-variance scaling empirically mitigates severe underestimation.

Related papers

• Annealed Softmax Greedy in Many-Armed Bayesian Bandits [9.553819152637493]
  Reinforcement learning with rewards (RLVR) and group-based policy optimization methods such as GRPO update a verifiable policy by sampling multiple completions per prompt.<n>This paper studies a stylized explanation for why uncertainty-agnostic updates can nevertheless be effective.
  arXiv  Detail & Related papers  (2026-05-29T09:05:29Z)
• How Neural Reward Models Learn Features for Policy Optimization: A Single-Index Analysis [53.063298916923976]
  We study the feedback in a Gaussian single-index model with $r*(x) = *(langle *, xrangle)$ and $x sim N(0, I_d)$.<n>We analyze a two-stage neural reward model that first learns the hidden direction $*$ from reward-weighted samples and then fits the readout layer by weighted ridge regression.
  arXiv  Detail & Related papers  (2026-05-23T22:00:38Z)
• Theoretical Limits of Language Model Alignment [9.45142272392467]
  Language model (LM) alignment improves model outputs to reflect human preferences while preserving the capabilities of the base model.<n>The most common alignment approaches are (i) reinforcement learning, which maximizes the expected reward under a KL-divergence constraint, and (ii) best-of-$N$ alignment, which selects the highest-reward output among $N$ independent samples.<n>We characterize the information-theoretic limits of KL-regularized alignment by deriving the maximum achievable expected reward gain for a fixed KL-divergence budget.
  arXiv  Detail & Related papers  (2026-05-08T01:32:22Z)
• Reinforcement Learning from Multi-Source Imperfect Preferences: Best-of-Both-Regimes Regret [71.69884486156359]
  We study episodic RL from emphmulti-source imperfect preferences through a cumulative imperfection budget.<n>We propose a unified algorithm with regret $tildeO(sqrtK/M+)$, which exhibits a best-of-both-regimes behavior.
  arXiv  Detail & Related papers  (2026-03-20T19:34:53Z)
• Decision from Suboptimal Classifiers: Excess Risk Pre- and Post-Calibration [52.70324949884702]
  We quantify the excess risk incurred using approximate posterior probabilities in batch binary decision-making.<n>We identify regimes where recalibration alone addresses most of the regret, and regimes where the regret is dominated by the grouping loss.<n>On NLP experiments, we show that these quantities identify when the expected gain of more advanced post-training is worth the operational cost.
  arXiv  Detail & Related papers  (2025-03-23T10:52:36Z)
• Continuous K-Max Bandits [54.21533414838677]
  We study the $K$-Max multi-armed bandits problem with continuous outcome distributions and weak value-index feedback.<n>This setting captures critical applications in recommendation systems, distributed computing, server scheduling, etc.<n>Our key contribution is the computationally efficient algorithm DCK-UCB, which combines adaptive discretization with bias-corrected confidence bounds.
  arXiv  Detail & Related papers  (2025-02-19T06:37:37Z)
• Combinatorial Bandits for Maximum Value Reward Function under Max Value-Index Feedback [9.771002043127728]
  We consider a multi-armed bandit problem for maximum value reward function under maximum value and index feedback. We propose an algorithm and provide a regret bound for problem instances with arm outcomes according to arbitrary distributions with finite supports. Our algorithm achieves a $O((k/Delta)log(T))$ distribution-dependent and a $tildeO(sqrtT)$ distribution-independent regret.
  arXiv  Detail & Related papers  (2023-05-25T14:02:12Z)
• Finite-Time Regret of Thompson Sampling Algorithms for Exponential Family Multi-Armed Bandits [88.21288104408556]
  We study the regret of Thompson sampling (TS) algorithms for exponential family bandits, where the reward distribution is from a one-dimensional exponential family. We propose a Thompson sampling, termed Expulli, which uses a novel sampling distribution to avoid the under-estimation of the optimal arm.
  arXiv  Detail & Related papers  (2022-06-07T18:08:21Z)
• Online nonparametric regression with Sobolev kernels [99.12817345416846]
  We derive the regret upper bounds on the classes of Sobolev spaces $W_pbeta(mathcalX)$, $pgeq 2, beta>fracdp$. The upper bounds are supported by the minimax regret analysis, which reveals that in the cases $beta> fracd2$ or $p=infty$ these rates are (essentially) optimal.
  arXiv  Detail & Related papers  (2021-02-06T15:05:14Z)
• Optimal Algorithms for Stochastic Multi-Armed Bandits with Heavy Tailed Rewards [24.983866845065926]
  We consider multi-armed bandits with heavy-tailed rewards, whose $p$-th moment is bounded by a constant $nu_p$ for $1pleq2$. We propose a novel robust estimator which does not require $nu_p$ as prior information. We show that an error probability of the proposed estimator decays exponentially fast.
  arXiv  Detail & Related papers  (2020-10-24T10:44:02Z)
• Lenient Regret for Multi-Armed Bandits [72.56064196252498]
  We consider the Multi-Armed Bandit (MAB) problem, where an agent sequentially chooses actions and observes rewards for the actions it took. While the majority of algorithms try to minimize the regret, i.e., the cumulative difference between the reward of the best action and the agent's action, this criterion might lead to undesirable results. We suggest a new, more lenient, regret criterion that ignores suboptimality gaps smaller than some $epsilon$.
  arXiv  Detail & Related papers  (2020-08-10T08:30:52Z)
• A General Theory of the Stochastic Linear Bandit and Its Applications [8.071506311915398]
  We introduce a general analysis framework and a family of algorithms for the linear bandit problem. Our new notion of optimism in expectation gives rise to a new algorithm, called sieved greedy (SG) that reduces the overexploration problem in OFUL. In addition to proving that SG is theoretically rate optimal, our empirical simulations show that SG outperforms existing benchmarks such as greedy, OFUL, and TS.
  arXiv  Detail & Related papers  (2020-02-12T18:54:41Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.