Theory · statistics papers, 2023–2026

Key findings

• Theory · statistics went from 16 papers in 2023 to 57 in 2026, roughly 3.6x in three years.
• NeurIPS publishes the most Theory · statistics work: 58 papers from 2023 to 2026, 42% of the four-venue total.
• The fastest-growing overlap is with Theory · learning theory: papers tagged with both rose from 1 in 2023 to 28 in 2026.
• Among the 5 topics under Theory and Statistics, it ranks #5 by 2026 paper count; the fastest grower in the group, Theory · learning theory, is up 5.6x versus 3.6x here.

Representative papers

Spread across years and venues so every edition shows up

• NeurIPS 2026Asymptotic Anytime-Valid Inference for U-statistics
• ICML 2026A Theory of How Pretraining Shapes Inductive Bias in Fine-Tuning
• ICLR 2026The Gaussian-Head OFL Family: One-Shot Federated Learning from Client Global Statistics
• NeurIPS 2025A Statistical Theory of Contrastive Learning via Approximate Sufficient Statistics
• ICML 2025Contextual Bandits for Unbounded Context Distributions
• ICLR 2025Improving Pretraining Data Using Perplexity Correlations
• CVPR 2025Dynamic Group Normalization: Spatio-Temporal Adaptation to Evolving Data Statistics
• NeurIPS 2024FedGTST: Boosting Global Transferability of Federated Models via Statistics Tuning
• ICML 2024Neural Networks Learn Statistics of Increasing Complexity
• ICLR 2024Kernelised Normalising Flows
• NeurIPS 2023Fast Online Changepoint Detection via Functional Pruning CUSUM Statistics
• ICML 2023Fair Densities via Boosting the Sufficient Statistics of Exponential Families
• ICLR 2023Don’t fear the unlabelled: safe semi-supervised learning via debiasing
• NeurIPS 2026Computationally sufficient statistics for Ising models
• ICML 2026Adaptive Multiscale Binary Expansion Tests for Independence
• ICLR 2026Bayesian Influence Functions for Hessian-Free Data Attribution
• NeurIPS 2025Backward Conformal Prediction
• ICML 2025Eigen Analysis of Conjugate Kernel and Neural Tangent Kernel
• ICLR 2025From Risk to Uncertainty: Generating Predictive Uncertainty Measures via Bayesian Estimation
• NeurIPS 2024Statistical Multicriteria Benchmarking via the GSD-Front

All Theory · statistics papers

138 papers, page 1 of 3

1. ICLR 2026A Law of Data Reconstruction for Random Features (And Beyond)
2. ICLR 2026Bayesian Influence Functions for Hessian-Free Data Attribution
3. ICLR 2026Beyond Spectra: Eigenvector Overlaps in Loss Geometry
4. ICLR 2026CaRe-BN: Precise Moving Statistics for Stabilizing Spiking Neural Networks in Reinforcement Learning
5. ICLR 2026Gradient Descent Dynamics of Rank-One Matrix Denoising
6. ICLR 2026High-Dimensional Analysis of Single-Layer Attention for Sparse-Token Classification
7. ICLR 2026Improved high-dimensional estimation with Langevin dynamics and stochastic weight averaging
8. ICLR 2026Learning the Inverse Temperature of Ising Models under Hard Constraints using One Sample
9. ICLR 2026Mean Estimation from Coarse Data: Characterizations and Efficient Algorithms
10. ICLR 2026Memory-Statistics Tradeoff in Continual Learning with Structural Regularization
11. ICLR 2026Predicting Kernel Regression Learning Curves from Only Raw Data Statistics
12. ICLR 2026Price of Quality: Sufficient Conditions for Sparse Recovery using Mixed-Quality Data
13. ICLR 2026Scaling Laws and Spectra of Shallow Neural Networks in the Feature Learning Regime
14. ICLR 2026Statistical Advantage of Softmax Attention: Insights from Single-Location Regression
15. ICLR 2026The Gaussian-Head OFL Family: One-Shot Federated Learning from Client Global Statistics
16. ICML 2026A Random Matrix Perspective on the Consistency of Diffusion Models
17. ICML 2026A Refined Generalization Analysis for Extreme Multi-class Supervised Contrastive Representation Learning
18. ICML 2026A Studentized Spherical Harmonics–Based Nonparametric Two-Sample Test for Compositional and Directional Data
19. ICML 2026A Theory of Contrastive Learning with Natural Images
20. ICML 2026A Theory of How Pretraining Shapes Inductive Bias in Fine-Tuning
21. ICML 2026A theory of learning data statistics in diffusion models, from easy to hard
22. ICML 2026Adaptive Multiscale Binary Expansion Tests for Independence
23. ICML 2026Approximating f -Divergences with Rank Statistics
24. ICML 2026Beyond First-order Asymptotics in Sequential Mean Testing
25. ICML 2026Characterization of Gaussian Universality Breakdown in High-Dimensional Empirical Risk Minimization
26. ICML 2026Combinatorial Sparse PCA Beyond the Spiked Identity Model
27. ICML 2026Deep networks learn to parse uniform-depth context-free languages from local statistics
28. ICML 2026Deep Neural Network Regression with Functional Covariates
29. ICML 2026Deriving Neural Scaling Laws from the Statistics of Natural Language
30. ICML 2026Finite-Width Neural Tangent Kernels from Feynman Diagrams
31. ICML 2026Learning $U$-Statistics with Active Inference
32. ICML 2026Learning Gaussian Graphical Models from a Glauber Trajectory Without Mixing
33. ICML 2026Linear Regression with Unknown Truncation Beyond Gaussian Features
34. ICML 2026Mean-Shift PCA by Knockoff Mean
35. ICML 2026On the Existence of Consistent Adversarial Attacks in High-Dimensional Linear Classification
36. ICML 2026On the Power of Statistics in Class-Incremental Learning with Pretrained Models
37. ICML 2026Optimal Attention Temperature Improves the Robustness of In-Context Learning under Distribution Shift in High Dimensions
38. ICML 2026Outcome-Aware Spectral Feature Learning for Instrumental Variable Regression
39. ICML 2026Provable Bounds for the Learnability of Sample-Compressible Families from Noisy Samples
40. ICML 2026Sample Complexity Bounds for Robust Mean Estimation with Mean-Shift Contamination
41. ICML 2026Sharp Concentration Bounds for Bundle-Valued Statistics on Manifolds
42. ICML 2026Sharp description of local minima in the loss landscape of high-dimensional two-layer ReLU neural networks
43. ICML 2026Symmetries in language statistics shape the geometry of model representations
44. ICML 2026The Catastrophic Failure of *the* k-Means Algorithm in High Dimensions, and How Hartigan's Algorithm Avoids It
45. ICML 2026When Softmax Fails at the Top: Extreme‑Value Corrections for InfoNCE
46. NeurIPS 2026Asymptotic Anytime-Valid Inference for U-statistics
47. NeurIPS 2026Computationally sufficient statistics for Ising models
48. NeurIPS 2026Hierarchical Concept Geometry in Language Representations Emerges from Word Co-occurrence
49. NeurIPS 2026Neural population tuning statistics as priors for multitask generalization
50. NeurIPS 2026Pretraining Data Statistics Shape the Phases of Learning Entity Comparison in Language Models

FAQ

How many Theory · statistics papers were published at NeurIPS, ICLR, ICML and CVPR from 2023 to 2026?

138 in total: 16 in 2023, 21 in 2024, 44 in 2025, 57 in 2026.

How many Theory · statistics papers are at NeurIPS 2026?

NeurIPS 2026 has 12 Theory · statistics papers, versus 23 the year before.

How many Theory · statistics papers are at ICLR 2026?

ICLR 2026 has 15 Theory · statistics papers, versus 7 the year before.

How many Theory · statistics papers are at ICML 2026?

ICML 2026 has 30 Theory · statistics papers, versus 13 the year before.

How many Theory · statistics papers are at CVPR 2025?

CVPR 2025 has 1 Theory · statistics papers.

Which conference publishes the most Theory · statistics papers?

NeurIPS, with 58 papers from 2023 to 2026.