Research map / Learning theory and algorithms / Kernels and regression theory
Statistical perspectives on generalization
Analyze generalization via invariance, interpolation, shift and information-theoretic views of data.
99 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Kernels and regression theory, keyword cluster “generalization bounds · pac · generalization error”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
31 papers in 2023–24 and 33 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 12.5% to 13.4%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2017 · Model-Powered Conditional Independence Test (NeurIPS 2017 · 41 citations)
- 2018 · Nonoverlap-Promoting Variable Selection (ICML 2018 · 6 citations)
- 2019 · Analogies Explained: Towards Understanding Word Embeddings (ICML 2019 · 54 citations)
- 2020 · A Closer Look at the Training Strategy for Modern Meta-Learning (NeurIPS 2020 · 15 citations)
- 2021 · Provably Strict Generalisation Benefit for Equivariant Models (ICML 2021 · 10 citations)
- 2022 · Model Agnostic Sample Reweighting for Out-of-Distribution Learning (ICML 2022 · 9 citations)
- 2023 · Stable Learning via Sparse Variable Independence (AAAI 2023 · 12 citations)
- 2024 · Restoring balance: principled under/oversampling of data for optimal classification (ICML 2024 · 5 citations)
- 2025 · Subsampled Ensemble Can Improve Generalization Tail Exponentially (NeurIPS 2025 · 2 citations)
- 2026 · Generalizing Analogical Inference from Boolean to Continuous Domains (AAAI 2026 · 1 citation)
Earliest papers here
- Model-Powered Conditional Independence Test (NeurIPS 2017 · 41 citations)
- Consistent Multitask Learning with Nonlinear Output Relations (NeurIPS 2017 · 23 citations)
- Learning Latent Space Models with Angular Constraints (ICML 2017 · 17 citations)
What has been tried
71 papers here report 553 measured changes (a component added, removed or swapped, compared with a baseline): helped 69%, no clear effect 8%, hurt 18%, mixed 5%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add LSPE layers to OAP | ZINC · MSE | \textbf{0.118 ± 0.0006} -> \textbf{0.098 ± 0.0009} | Helped | A Canonicalization Perspective on Invariant and Equivariant Learning (NeurIPS 2024 · 5 citations) |
| Use ROVE instead of ROVEs (do not split the data between phases) | Stochastic programs · expected out-of-sample cost | Helped | Subsampled Ensemble Can Improve Generalization Tail Exponentially (NeurIPS 2025 · 2 citations) | |
| Add de-biasing by estimating task weights and representation gradients on independent data batches | Non-isotropic linear representation learning · representation-update noise | Helped | Sample-Efficient Linear Representation Learning from Non-IID Non-Isotropic Data (ICLR 2024 · 2 citations) | |
| use log propensity mapping instead of power mapping | COAT · AUC | 0.7710±0.0014 -> 0.7804±0.0023 | Helped | Fine-Grained Dynamic Framework for Bias-Variance Joint Optimization on Data Missing Not at Random (NeurIPS 2024 · 2 citations) |
| Use a smaller sketching dimension m | StanfordCars, low data regime · finetuning performance | Helped | Sketchy Moment Matching: Toward Fast and Provable Data Selection for Finetuning (NeurIPS 2024 · 1 citation) | |
| apply batch balancing | COMPAS gender, n=250 · DP | 0.056 -> 0.042 | Helped | Fairness Overfitting in Machine Learning: An Information-Theoretic Perspective (ICML 2025 · 0 citations) |
| Select K by cross-validation instead of setting K=√n | Simulated Examples 2 and 3 · AM-risk | Helped | Concentration and excess risk bounds for imbalanced classification with synthetic oversampling (NeurIPS 2025 · 0 citations) | |
| Penalize search with the PAC-Bayes complexity term (SRM) | CIFAR-10 · generalization bound | Helped | Understanding prompt engineering may not require rethinking generalization (ICLR 2024 · 0 citations) | |
| Increase the number of folds in M-fold holdout | HPO benchmark experiments · effect of reshuffling on generalization performance | Hurt | Reshuffling Resampling Splits Can Improve Generalization of Hyperparameter Optimization (NeurIPS 2024 · 0 citations) | |
| Use a learnable rather than fixed temperature τ | DFN-12M · ImageNet Top 1 accuracy | 37.65 -> 46.31 | Helped | Model Steering: Learning with a Reference Model Improves Generalization Bounds and Scaling Laws (ICML 2025) |
What reviewers will expect
Datasets and baselines named in the 18 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- Adult · 2 of 18 papers
- California · 2 of 18 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 64 papers; from OpenAlex and the papers' affiliations).
- Google · 5 papers
- Columbia University · 4 papers
- Massachusetts Institute of Technology · 4 papers
- New York University · 3 papers
- Carnegie Mellon University · 3 papers
- University of California, Riverside · 2 papers
- Hong Kong University of Science and Technology · 2 papers
- Tsinghua University · 2 papers
- Weizmann Institute · 2 papers
- University of Pennsylvania · 2 papers
Use it
Download BibTeX of 39 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Information-theoretic generalization bounds (94 papers)
- Generalization theory of deep networks (67 papers)
- Empirical generalization measures (61 papers)
- PAC-Bayes bounds (39 papers)
Elsewhere in the topic
- RKHS kernel methods (127 papers)
- Neural tangent kernel analysis (109 papers)
- Kernel regression under shift (89 papers)
- High-dimensional asymptotic generalization (82 papers)
- High-dimensional linear regression theory (61 papers)
- Kernel ridge regression analysis (57 papers)
- Double descent and overfitting (53 papers)
- Random feature approximation (36 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).