Autoregressive Learning in Joint KL: Sharp Oracle Bounds and Lower Bounds

• URL: http://arxiv.org/abs/2605.12316v1
• Date: Tue, 12 May 2026 16:01:29 GMT
• Title: Autoregressive Learning in Joint KL: Sharp Oracle Bounds and Lower Bounds
• Abstract summary: We study the fundamental and timely problem of learning long sequences in autoregressive modeling and next-token prediction under model misspecification.<n>Our goal is to characterize how the sequence horizon (H) affects both approximation and estimation errors in this joint-distribution, sequence-level regime.
• Score: 8.164687789644365
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We study the fundamental and timely problem of learning long sequences in autoregressive modeling and next-token prediction under model misspecification, measured by the joint Kullback--Leibler (KL) divergence. Our goal is to characterize how the sequence horizon \(H\) affects both approximation and estimation errors in this joint-distribution, sequence-level regime. By establishing matching upper and lower bounds, we provide, to our knowledge, the first complete characterization of long-horizon error behavior under the natural joint KL objective, with improved rates and optimality justification relative to existing work. On the approximation side, we show that joint KL admits a horizon-free approximation factor, in sharp contrast to Hellinger-based analyses that exhibit an \(Ω(H)\) dependence for computationally efficient methods; this isolates the choice of divergence as the source of approximation amplification. On the estimation side, we prove a fundamental information-theoretic lower bound of order \(Ω(H)\) that holds for both decomposable policy classes and fully shared policies, matching the \(\widetilde O(H)\) upper bounds achieved by computationally efficient algorithms. Our analysis clarifies the landscape of recent autoregressive learning results by aligning the log-loss training objective, the sequence-level evaluation metric, and the approximation metric {\color{black}through a sharp joint-KL oracle theory}. We further show that these joint-KL guarantees imply policy learning regret bounds at rates matching prior imitation learning literature.

Related papers

• Bridging the Gap Between Average and Discounted TD Learning [14.556544278062793]
  We introduce a novel algorithm designed explicitly for policy evaluation in the average-reward setting.<n>Our proposed method overcomes previous limitations by guaranteeing convergence to the unique solution of a properly defined Bellman equation.
  arXiv  Detail & Related papers  (2026-05-03T23:54:36Z)
• On Multi-Step Theorem Prediction via Non-Parametric Structural Priors [50.16583672681106]
  In this work, we explore training-free theorem prediction through the lens of in-context learning (ICL)<n>We propose Theorem Precedence Graphs, which encode temporal dependencies from historical solution traces as directed graphs, and impose explicit topological constraints that effectively prune the search space during inference.<n>Experiments on the FormalGeo7k benchmark show that our method achieves 89.29% accuracy, substantially outperforming ICL baselines and matching state-of-the-art supervised models.
  arXiv  Detail & Related papers  (2026-03-05T06:08:50Z)
• Uncertainty quantification for Markov chain induced martingales with application to temporal difference learning [55.197497603087065]
  We analyze the performance of the Temporal Difference (TD) learning algorithm with linear function approximations.<n>We establish novel and general high-dimensional concentration inequalities and Berry-Esseen bounds for vector-valued martingales induced by Markov chains.
  arXiv  Detail & Related papers  (2025-02-19T15:33:55Z)
• On Volume Minimization in Conformal Regression [8.673942897414934]
  We study the question of volume optimality in split conformal regression.<n>We first derive a finite-sample upper-bound on the excess volume loss of the interval returned by the classical split method.<n>We introduce EffOrt, a methodology that modifies the learning step so that the base prediction function is selected in order to minimize the length of the returned intervals.
  arXiv  Detail & Related papers  (2025-02-14T08:14:22Z)
• Full error analysis of policy gradient learning algorithms for exploratory linear quadratic mean-field control problem in continuous time with common noise [0.0]
  We study policy gradient (PG) learning and first demonstrate convergence in a model-based setting. We prove the global linear convergence and sample complexity of the PG algorithm with two-point gradient estimates in a model-free setting. In this setting, the parameterized optimal policies are learned from samples of the states and population distribution.
  arXiv  Detail & Related papers  (2024-08-05T14:11:51Z)
• Generalization Bounds of Surrogate Policies for Combinatorial Optimization Problems [53.03951222945921]
  We analyze smoothed (perturbed) policies, adding controlled random perturbations to the direction used by the linear oracle.<n>Our main contribution is a generalization bound that decomposes the excess risk into perturbation bias, statistical estimation error, and optimization error.<n>We illustrate the scope of the results on applications such as vehicle scheduling, highlighting how smoothing enables both tractable training and controlled generalization.
  arXiv  Detail & Related papers  (2024-07-24T12:00:30Z)
• Adaptive Federated Learning Over the Air [108.62635460744109]
  We propose a federated version of adaptive gradient methods, particularly AdaGrad and Adam, within the framework of over-the-air model training. Our analysis shows that the AdaGrad-based training algorithm converges to a stationary point at the rate of $mathcalO( ln(T) / T 1 - frac1alpha ).
  arXiv  Detail & Related papers  (2024-03-11T09:10:37Z)
• Horizon-Free and Instance-Dependent Regret Bounds for Reinforcement Learning with General Function Approximation [26.277745106128197]
  We propose an algorithm to tackle long planning horizon problems in reinforcement learning with general function approximation. The derived regret bound is deemed emphsharp as it matches the minimax lower bound when specialized to linear mixture MDPs up to logarithmic factors. The achievement of such a horizon-free, instance-dependent, and sharp regret bound hinges upon (i) novel algorithm designs and (ii) fine-grained analyses.
  arXiv  Detail & Related papers  (2023-12-07T17:35:34Z)
• Understanding Augmentation-based Self-Supervised Representation Learning via RKHS Approximation and Regression [53.15502562048627]
  Recent work has built the connection between self-supervised learning and the approximation of the top eigenspace of a graph Laplacian operator. This work delves into a statistical analysis of augmentation-based pretraining.
  arXiv  Detail & Related papers  (2023-06-01T15:18:55Z)
• Fine-grained analysis of non-parametric estimation for pairwise learning [9.676007573960383]
  We are concerned with the generalization performance of non-parametric estimation for pairwise learning. Our results can be used to handle a wide range of pairwise learning problems including ranking, AUC, pairwise regression and metric and similarity learning.
  arXiv  Detail & Related papers  (2023-05-31T08:13:14Z)
• Differentiable Annealed Importance Sampling and the Perils of Gradient Noise [68.44523807580438]
  Annealed importance sampling (AIS) and related algorithms are highly effective tools for marginal likelihood estimation. Differentiability is a desirable property as it would admit the possibility of optimizing marginal likelihood as an objective. We propose a differentiable algorithm by abandoning Metropolis-Hastings steps, which further unlocks mini-batch computation.
  arXiv  Detail & Related papers  (2021-07-21T17:10:14Z)

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.