←── back to feed
/topics/arxiv-machine-learning-papers-october-8

arXiv machine learning papers October 8

21 items●1 sources●updated 31d ago●trend 0

On October 8, 2026, arXiv published 20 machine learning papers spanning Bayesian optimization, federated learning, diffusion models, neural PDE solvers, and representation learning. Topics include optimizer batch-size dependence, offline RL with latent-space constraints, conformal prediction under covariate shift, and theoretical analyses of neural network training dynamics.

  • Bayesian optimization exploits evaluation latency by fitting full posterior over hyperparameters rather than point estimates (arXiv:2610.08969)
  • Optimizer performance varies with batch size; no universal scaling rule preserves best optimizer across training settings (arXiv:2610.08975)
  • Offline RL method LASER uses entropy regularization in latent space to prevent policy collapse and critic artifacts (arXiv:2610.08989)
  • Sketched calibration compresses covariates to reduce exponential cost of weighted conformal prediction under covariate shift (arXiv:2610.09208)
  • Two-layer linear networks achieve global exponential convergence with explicit rate in rich scaling regime (arXiv:2610.09356)
  • Individual world latents recoverable without reconstruction via distributional asymmetries in joint-embedding architectures (arXiv:2610.09457)
[BLG]blog/rss21
Work While They Sleep: Exploiting Evaluation Latency for Fully Bayesian Optimization
arXiv stat.ML · Gustavo Sutter, Alejandro Comas-Leon, David Holzm\"uller, Hao Wang, Luis Ricardez-Sandoval, Pascal Poupart, Agustinus Kristiadi · 31d
The Best Optimizer Depends on Batch Size
arXiv stat.ML · Xingyu Dang, Kaiyue Wen, Sadhika Malladi · 31d
LASER: Latent Space Adjoint Matching for Support-Constrained Entropy-Regularized Offline RL
arXiv stat.ML · Songyuan Zhang, Oswin So, Eric Yang Yu, Matthew Cleaveland, Peter Crowley-Dolen, Chuchu Fan · 31d
Graph-monotone entrywise guarantees for MLE and Rank Centrality on general comparison graphs
arXiv stat.ML · Yuepeng Yang, Cong Ma · 31d
FedRSPO+: A Heterogeneity-aware Algorithm for Decision-focused Federated Learning
arXiv stat.ML · Konstantinos Ziliaskopoulos, Alexander Vinel, Jiaqi Wang · 31d
Relative Wasserstein Spatial Depth for Cluster Number Selection in Distributional Data
arXiv stat.ML · Chenqin Xu, Yisha Yao · 31d
Lower Bounds for Parallel Diffusion Sampling
arXiv stat.ML · Yiwen Kou, Yimeng Wang · 31d
Exact Dynamics and Finite-Sample Trajectory Recovery of Linear Recursive Feature Machines
arXiv stat.ML · Andrew Cheng, Bobak T. Kiani, Yue M. Lu, Adityanarayanan Radhakrishnan · 31d
An Accuracy--Information Tradeoff for Loss-Difference Conditional Mutual Information
arXiv stat.ML · Hazar Yueksel · 31d
Sketched Calibration for Conformal Prediction under Covariate Shift
arXiv stat.ML · Mehrdad Pournaderi · 31d
Efficient Best-of-N policy evaluation for inference-time alignment
arXiv stat.ML · Jonas Schweisthal, Yuxin Wang, Athiya Deviyani, Stefan Feuerriegel, Dennis Frauen · 31d
The Symbol of the Surrogate: Measuring Numerical Provenance in Neural PDE Solvers
arXiv stat.ML · Ridham Patel · 31d
The Geometry of Anisotropic Dilation for Optimal Regularization
arXiv stat.ML · Carson Newman, Oscar Leong · 31d
Benign Overfitting under Heterogeneous Input Fusion
arXiv stat.ML · Houzhen Liu, Xiaobo Xia · 31d
Global Exponential Convergence of Two-Layer Linear Network Training
arXiv stat.ML · Stephen Y Zhang, Gabriel Peyr\'e · 31d
Finite-Rank Logistic Gaussian Processes with Exact Likelihood for Conditional Density Estimation
arXiv stat.ML · Jaehoan Kim, Indrajit Ghosh, Debdeep Pati, Dipankar Bandyopadhyay · 31d
DSReg: Provably Recovering Individual World Latents without Reconstruction
arXiv stat.ML · Yujia Zheng, David Klindt, Randall Balestriero, Bernhard Sch\"olkopf · 31d
Cluster-Robust Prediction-Powered Inference
arXiv stat.ML · David Broska, Michael Howes · 31d
Gauss-Newton Accuracy and Indefinite Hessians: Uniform Coexistence in Low-Cost Sets
arXiv stat.ML · Kihun Rhee, Hanjoon Byun · 31d
The Silhouette Operator: Identifiability of Low-Rank Measures from One-Dimensional Projections
arXiv stat.ML · Robert A. Vandermeulen · 31d
Unbounded Characteristic and Universal Kernels
arXiv stat.ML · Jose Cribeiro-Ramallo, Florian Kalinke, Zolt\'an Szab\'o · 31d