# No One Representation to Rule Them All: Overlapping Features of Training Methods

arXiv: https://arxiv.org/abs/2110.12899 · HTML: https://mlsanity.com/paper/2110.12899/

## Key facts

- Cited by 58 papers in the map (data through 14 Oct 2026).
- Builds on 48 of its 59 references that are in the map.
- First posted on arXiv 20 Oct 2021.

## Builds on (most cited first, 30 of 48)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Very Deep Convolutional Networks for Large-Scale Image Recognition](https://mlsanity.com/paper/1409.1556/) (2014)
- [A Simple Framework for Contrastive Learning of Visual Representations](https://mlsanity.com/paper/2002.05709/) (2020)
- [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks](https://mlsanity.com/paper/1905.11946/) (2019)
- [Rethinking the Inception Architecture for Computer Vision](https://mlsanity.com/paper/1512.00567/) (2015)
- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [MobileNetV2: Inverted Residuals and Linear Bottlenecks](https://mlsanity.com/paper/1801.04381/) (2018)
- [Squeeze-and-Excitation Networks](https://mlsanity.com/paper/1709.01507/) (2017)
- [Training data-efficient image transformers & distillation through attention](https://mlsanity.com/paper/2012.12877/) (2020)
- [Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision](https://mlsanity.com/paper/2102.05918/) (2021)
- [What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?](https://mlsanity.com/paper/1703.04977/) (2017)
- [RandAugment: Practical automated data augmentation with a reduced search space](https://mlsanity.com/paper/1909.13719/) (2019)
- [Do ImageNet Classifiers Generalize to ImageNet?](https://mlsanity.com/paper/1902.10811/) (2019)
- [ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness](https://mlsanity.com/paper/1811.12231/) (2018)
- [Deep High-Resolution Representation Learning for Visual Recognition](https://mlsanity.com/paper/1908.07919/) (2019)
- [Averaging Weights Leads to Wider Optima and Better Generalization](https://mlsanity.com/paper/1803.05407/) (2018)
- [Scaling Vision Transformers](https://mlsanity.com/paper/2106.04560/) (2021)
- [Adversarial Examples Are Not Bugs, They Are Features](https://mlsanity.com/paper/1905.02175/) (2019)
- [Revisiting Unreasonable Effectiveness of Data in Deep Learning Era](https://mlsanity.com/paper/1707.02968/) (2017)
- [Big Transfer (BiT): General Visual Representation Learning](https://mlsanity.com/paper/1912.11370/) (2019)
- [Diversity is All You Need: Learning Skills without a Reward Function](https://mlsanity.com/paper/1802.06070/) (2018)
- [AutoAugment: Learning Augmentation Policies from Data](https://mlsanity.com/paper/1805.09501/) (2018)
- [Do Better ImageNet Models Transfer Better?](https://mlsanity.com/paper/1805.08974/) (2018)
- [Underspecification Presents Challenges for Credibility in Modern Machine Learning](https://mlsanity.com/paper/2011.03395/) (2020)
- [Deep Ensembles: A Loss Landscape Perspective](https://mlsanity.com/paper/1912.02757/) (2019)
- [Deep Layer Aggregation](https://mlsanity.com/paper/1707.06484/) (2017)
- [Selective Kernel Networks](https://mlsanity.com/paper/1903.06586/) (2019)
- [Meta Pseudo Labels](https://mlsanity.com/paper/2003.10580/) (2020)

## Built on by (most cited first, 30 of 58)

- [Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time](https://mlsanity.com/paper/2203.05482/) (2022)
- [Branch-Train-Merge: Embarrassingly Parallel Training of Expert Language Models](https://mlsanity.com/paper/2208.03306/) (2022)
- [Diverse Weight Averaging for Out-of-Distribution Generalization](https://mlsanity.com/paper/2205.09739/) (2022)
- [WARM: On the Benefits of Weight Averaged Reward Models](https://mlsanity.com/paper/2401.12187/) (2024)
- [Quality Not Quantity: On the Interaction between Dataset Design and Robustness of CLIP](https://mlsanity.com/paper/2208.05516/) (2022)
- [Model Ratatouille: Recycling Diverse Models for Out-of-Distribution Generalization](https://mlsanity.com/paper/2212.10445/) (2022)
- [Head2Toe: Utilizing Intermediate Representations for Better Transfer Learning](https://mlsanity.com/paper/2201.03529/) (2022)
- [Accelerating Dataset Distillation via Model Augmentation](https://mlsanity.com/paper/2212.06152/) (2022)
- [When does dough become a bagel? Analyzing the remaining mistakes on ImageNet](https://mlsanity.com/paper/2205.04596/) (2022)
- [Language Models in the Loop: Incorporating Prompting into Weak Supervision](https://mlsanity.com/paper/2205.02318/) (2022)
- [Compute-Efficient Deep Learning: Algorithmic Trends and Opportunities](https://mlsanity.com/paper/2210.06640/) (2022)
- [Understanding Cross-Domain Few-Shot Learning Based on Domain Similarity and Few-Shot Difficulty](https://mlsanity.com/paper/2202.01339/) (2022)
- [Tangent Model Composition for Ensembling and Continual Fine-tuning](https://mlsanity.com/paper/2307.08114/) (2023)
- [Downstream Datasets Make Surprisingly Good Pretraining Corpora](https://mlsanity.com/paper/2209.14389/) (2022)
- [Deconstructing Distributions: A Pointwise Framework of Learning](https://mlsanity.com/paper/2202.09931/) (2022)
- [lo-fi: distributed fine-tuning without communication](https://mlsanity.com/paper/2210.11948/) (2022)
- [Weighted Ensemble Self-Supervised Learning](https://mlsanity.com/paper/2211.09981/) (2022)
- [Fantastic Gains and Where to Find Them: On the Existence and Prospect of General Knowledge Transfer between Any Pretrained Model](https://mlsanity.com/paper/2310.17653/) (2023)
- [Revisiting adapters with adversarial training](https://mlsanity.com/paper/2210.04886/) (2022)
- [Pathologies of Predictive Diversity in Deep Ensembles](https://mlsanity.com/paper/2302.00704/) (2023)
- [Domain Aligned CLIP for Few-shot Classification](https://mlsanity.com/paper/2311.09191/) (2023)
- [Agreement-Based Cascading for Efficient Inference](https://mlsanity.com/paper/2407.02348/) (2024)
- [Lessons and Insights from a Unifying Study of Parameter-Efficient Fine-Tuning (PEFT) in Visual Recognition](https://mlsanity.com/paper/2409.16434/) (2024)
- [Learning useful representations for shifting tasks and distributions](https://mlsanity.com/paper/2212.07346/) (2022)
- [Explore and Exploit the Diverse Knowledge in Model Zoo for Domain Generalization](https://mlsanity.com/paper/2306.02595/) (2023)
- [Towards Inference Efficient Deep Ensemble Learning](https://mlsanity.com/paper/2301.12378/) (2023)
- [A Holistic Assessment of the Reliability of Machine Learning Systems](https://mlsanity.com/paper/2307.10586/) (2023)
- [Scaling (Down) CLIP: A Comprehensive Analysis of Data, Architecture, and Training Strategies](https://mlsanity.com/paper/2404.08197/) (2024)
- [WASH: Train your Ensemble with Communication-Efficient Weight Shuffling, then Average](https://mlsanity.com/paper/2405.17517/) (2024)
- [Post-Hoc Reversal: Are We Selecting Models Prematurely?](https://mlsanity.com/paper/2404.07815/) (2024)

## Abstract (arXiv)

Despite being able to capture a range of features of the data, high accuracy models trained with supervision tend to make similar predictions. This seemingly implies that high-performing models share similar biases regardless of training methodology, which would limit ensembling benefits and render low-accuracy models as having little practical use. Against this backdrop, recent work has developed quite different training techniques, such as large-scale contrastive learning, yielding competitively high accuracy on generalization and robustness benchmarks. This motivates us to revisit the assumption that models necessarily learn similar functions. We conduct a large-scale empirical study of models across hyper-parameters, architectures, frameworks, and datasets. We find that model pairs that diverge more in training methodology display categorically different generalization behavior, producing increasingly uncorrelated errors. We show these models specialize in subdomains of the data, leading to higher ensemble performance: with just 2 models (each with ImageNet accuracy ~76.5%), we can create ensembles with 83.4% (+7% boost). Surprisingly, we find that even significantly low-accuracy models can be used to improve high-accuracy models. Finally, we show diverging training methodology yield representations that capture overlapping (but not supersetting) feature sets which, when combined, lead to increased downstream performance.