mlsanityData to 7 Oct 2026

Datamodels: Predicting Predictions from Training Data

arXiv 2202.00622 ↗First posted 1 Feb 2022stat.ML cs.CV cs.LGcited by 186

Not on the map. Why?

Abstract (arXiv)

We present a conceptual framework, datamodeling, for analyzing the behavior of a model class in terms of the training data. For any fixed "target" example x, training set S, and learning algorithm, a datamodel is a parameterized function 2S → ℝ that for any subset of S' subset S -- using only information about which examples of S are contained in S' -- predicts the outcome of training a model on S' and evaluating on x. Despite the potential complexity of the underlying process being approximated (e.g., end-to-end training and evaluation of deep neural networks), we show that even simple linear datamodels can successfully predict model outputs. We then demonstrate that datamodels give rise to a variety of applications, such as: accurately predicting the effect of dataset counterfactuals; identifying brittle predictions; finding semantically similar examples; quantifying train-test leakage; and embedding data into a well-behaved and feature-rich representation space. Data for this paper (including pre-computed datamodels as well as raw predictions from four million trained deep neural networks) is available at https://github.com/MadryLab/datamodels-data .

Key facts. Cited by 186 papers in the map (data through 7 Oct 2026). Builds on 53 of its 98 references that are in the map. First posted on arXiv 1 Feb 2022.

Builds on

53 of 98 references.

References · 53

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  3. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  4. 2017A Unified Approach to Interpreting Model Predictionscited by 6,488
  5. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,040
  6. 2016Simple and Scalable Predictive Uncertainty Estimation using Deep Ensemblescited by 3,888
  7. 2016Deep Learning with Differential Privacycited by 2,682
  8. 2020Extracting Training Data from Large Language Modelscited by 2,144
  9. 2017Understanding Black-box Predictions via Influence Functionscited by 1,714
  10. 2019Similarity of Neural Network Representations Revisitedcited by 1,695
  11. 2016Understanding deep learning requires rethinking generalizationcited by 1,657
  12. 2017Improved Regularization of Convolutional Neural Networks with Cutoutcited by 1,623
38 more
  1. 2020WILDS: A Benchmark of in-the-Wild Distribution Shiftscited by 1,366
  2. 2017BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chaincited by 1,052
  3. 2019Adversarial Examples Are Not Bugs, They Are Featurescited by 950
  4. 2021Deduplicating Training Data Makes Language Models Bettercited by 705
  5. 2018An Empirical Study of Example Forgetting during Deep Neural Network Learningcited by 669
  6. 2019Data Shapley: Equitable Valuation of Data for Machine Learningcited by 617
  7. 2020Estimating Training Data Influence by Tracing Gradient Descentcited by 565
  8. 2020What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimationcited by 492
  9. 2020Underspecification Presents Challenges for Credibility in Modern Machine Learningcited by 488
  10. 2019Does Learning Require Memorization? A Short Tale about a Long Tailcited by 452
  11. 2017Functional Map of the Worldcited by 338
  12. 2019Selection via Proxy: Efficient Data Selection for Deep Learningcited by 337
  13. 2019Differential Privacy Has Disparate Impact on Model Accuracycited by 316
  14. 2020An Investigation of Why Overparameterization Exacerbates Spurious Correlationscited by 310
  15. 2019Towards Efficient Data Valuation Based on the Shapley Valuecited by 279
  16. 2020Influence Functions in Deep Learning Are Fragilecited by 247
  17. 2021Deep learning: a statistical viewpointcited by 239
  18. 2021Revisiting Model Stitching to Compare Neural Representationscited by 203
  19. 2018Representer Point Selection for Explaining Deep Neural Networkscited by 187
  20. 2019On the Accuracy of Influence Functions for Measuring Group Effectscited by 146
  21. 2021Assessing Generalization of SGD via Disagreementcited by 116
  22. 2020Certified Robustness to Label-Flipping Attacks via Randomized Smoothingcited by 103
  23. 2020When is Memorization of Irrelevant Training Data Necessary for High-Accuracy Learning?cited by 102
  24. 2021The MultiBERTs: BERT Reproductions for Robustness Analysiscited by 89
  25. 2021Leveraging Sparse Linear Layers for Debuggable Deep Networkscited by 82
  26. 2021Input Similarity from the Neural Network Perspectivecited by 78
  27. 2019Do We Train on Test Data? Purging CIFAR of Near-Duplicatescited by 61
  28. 2020Distributional Generalization: A New Kind of Generalizationcited by 41
  29. 2019Model Similarity Mitigates Test Set Overusecited by 37
  30. 2021An Empirical Comparison of Instance Attribution Methods for NLPcited by 35
  31. 2021Are Larger Pretrained Language Models Uniformly Better? Comparing Performance at the Instance Levelcited by 34
  32. 2019Identity Crisis: Memorization and Generalization under Extreme Overparameterizationcited by 33
  33. 2018Celer: a Fast Solver for the Lasso with Dual Extrapolationcited by 31
  34. 20213DB: A Framework for Debugging Computer Vision Modelscited by 31
  35. 2019bLIMEy: Surrogate Prediction Explanations Beyond LIMEcited by 17
  36. 2020A Corrective View of Neural Networks: Representation, Memorization and Learningcited by 13
  37. 2016Prediction models for network-linked datacited by 12
  38. 2019Regression from Dependent Observationscited by 10

50 most cited of 53 All 53 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 20; 2023: 32; 2024: 54; 2025: 49; 2026: 31.

186 papers.

Sort:

By area:

  1. 2023DataComp: In search of the next generation of multimodal datasetscited by 635
  2. 2024LESS: Selecting Influential Data for Targeted Instruction Tuningcited by 514
  3. 2023SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generationcited by 376
  4. 2023TRAK: Attributing Model Behavior at Scalecited by 273
  5. 2022The Privacy Onion Effect: Memorization is Relativecited by 128
  6. 2023Similarity of Neural Network Models: A Survey of Functional and Representational Measurescited by 128
  7. 2022Data Banzhaf: A Robust Data Valuation Framework for Machine Learningcited by 126
  8. 2023Skill-it! A Data-Driven Skills Framework for Understanding and Training Language Modelscited by 120
  9. 2023Foundation Models and Fair Usecited by 112
  10. 2024DsDm: Model-Aware Dataset Selection with Datamodelscited by 103
  11. 2024MATES: Model-Aware Data Selection for Efficient Pretraining with Data Influence Modelscited by 96
  12. 2024ContextCite: Attributing Model Generation to Contextcited by 93
38 more
  1. 2022Distilling Model Failures as Directions in Latent Spacecited by 82
  2. 2022Cramming: Training a Language Model on a Single GPU in One Daycited by 81
  3. 2024Data Shapley in One Training Runcited by 76
  4. 2023In-context Example Selection with Influencescited by 60
  5. 2022Data Curation Alone Can Stabilize In-context Learningcited by 58
  6. 2022Measuring the Effect of Training Data on Deep Learning Predictions via Randomized Experimentscited by 50
  7. 2023Intriguing Properties of Data Attribution on Diffusion Modelscited by 47
  8. 2024Improving Pretraining Data Using Perplexity Correlationscited by 46
  9. 2024Incremental Residual Concept Bottleneck Modelscited by 43
  10. 2023The Journey, Not the Destination: How Data Guides Diffusion Modelscited by 42
  11. 2024Scalable Influence and Fact Tracing for Large Language Model Pretrainingcited by 40
  12. 2022Canary in a Coalmine: Better Membership Inference with Ensembled Adversarial Queriescited by 36
  13. 2024Capturing the Temporal Dependence of Training Data Influencecited by 34
  14. 2024Training Data Attribution via Approximate Unrolled Differentiationcited by 33
  15. 2022A Data-Based Perspective on Transfer Learningcited by 32
  16. 2024Most Influential Subset Selection: Challenges, Promises, and Beyondcited by 31
  17. 2022ModelDiff: A Framework for Comparing Learning Algorithmscited by 30
  18. 2024Rethinking Data Shapley for Data Selection Tasks: Misleads and Meritscited by 29
  19. 2025MoDoMoDo: Multi-Domain Data Mixtures for Multimodal LLM Reinforcement Learningcited by 29
  20. 2022Exploring Transformer Backbones for Heterogeneous Treatment Effect Estimationcited by 25
  21. 2024Hybrid Preferences: Learning to Route Instances for Human vs. AI Feedbackcited by 25
  22. 2025DataMIL: Selecting Data for Robot Imitation Learning with Datamodelscited by 25
  23. 2022Understanding Influence Functions and Datamodels via Harmonic Analysiscited by 23
  24. 2023Exploring the Limits of Model-Targeted Indiscriminate Data Poisoning Attackscited by 23
  25. 2024Data Attribution for Text-to-Image Models by Unlearning Synthesized Imagescited by 22
  26. 2023Internet Explorer: Targeted Representation Learning on the Open Webcited by 21
  27. 2024Stochastic Amortization: A Unified Approach to Accelerate Feature and Data Attributioncited by 21
  28. 2024Diffusion Attribution Score: Evaluating Training Data Influence in Diffusion Modelscited by 21
  29. 2023Neural Relation Graph: A Unified Framework for Identifying Label Noise and Outlier Datacited by 20
  30. 2023Identification of Negative Transfers in Multitask Learning Using Surrogate Modelscited by 20
  31. 2023On the Variance of Neural Network Training with respect to Test Sets and Distributionscited by 20
  32. 2024CoLoR-Filter: Conditional Loss Reduction Filtering for Targeted Language Model Pre-trainingcited by 20
  33. 2024Scalable Multitask Learning Using Gradient-based Estimation of Task Affinitycited by 20
  34. 2025MAGIC: Near-Optimal Data Attribution for Deep Learningcited by 20
  35. 2024Decomposing and Editing Predictions by Modeling Model Computationcited by 19
  36. 2024Task-Adaptive Pretrained Language Models via Clustered-Importance Samplingcited by 18
  37. 2024textttdattri: A Library for Efficient Data Attributioncited by 18
  38. 2024Attribute-to-Delete: Machine Unlearning via Datamodel Matchingcited by 18

50 most cited of 186 All 186 (JSON)