No "Zero-Shot" Without Exponential Data: Pretraining Concept Frequency Determines Multimodal Model Performance

• URL: http://arxiv.org/abs/2404.04125v3
• Date: Tue, 29 Oct 2024 14:00:46 GMT
• Title: No "Zero-Shot" Without Exponential Data: Pretraining Concept Frequency Determines Multimodal Model Performance
• Authors: Vishaal Udandarao, Ameya Prabhu, Adhiraj Ghosh, Yash Sharma, Philip H. S. Torr, Adel Bibi, Samuel Albanie, Matthias Bethge,
• Abstract summary: multimodal models require exponentially more data to achieve linear improvements in downstream "zero-shot" performance. Our study reveals an exponential need for training data which implies that the key to "zero-shot" generalization capabilities under large-scale training paradigms remains to be found.
• Score: 68.18779562801762
• License:
• Abstract: Web-crawled pretraining datasets underlie the impressive "zero-shot" evaluation performance of multimodal models, such as CLIP for classification/retrieval and Stable-Diffusion for image generation. However, it is unclear how meaningful the notion of "zero-shot" generalization is for such multimodal models, as it is not known to what extent their pretraining datasets encompass the downstream concepts targeted for during "zero-shot" evaluation. In this work, we ask: How is the performance of multimodal models on downstream concepts influenced by the frequency of these concepts in their pretraining datasets? We comprehensively investigate this question across 34 models and five standard pretraining datasets (CC-3M, CC-12M, YFCC-15M, LAION-400M, LAION-Aesthetics), generating over 300GB of data artifacts. We consistently find that, far from exhibiting "zero-shot" generalization, multimodal models require exponentially more data to achieve linear improvements in downstream "zero-shot" performance, following a sample inefficient log-linear scaling trend. This trend persists even when controlling for sample-level similarity between pretraining and downstream datasets, and testing on purely synthetic data distributions. Furthermore, upon benchmarking models on long-tailed data sampled based on our analysis, we demonstrate that multimodal models across the board perform poorly. We contribute this long-tail test set as the "Let it Wag!" benchmark to further research in this direction. Taken together, our study reveals an exponential need for training data which implies that the key to "zero-shot" generalization capabilities under large-scale training paradigms remains to be found.

Related papers

• A Mamba Foundation Model for Time Series Forecasting [13.593170999506889]
  We introduce TSMamba, a linear-complexity foundation model for time series forecasting built on the Mamba architecture. The model captures temporal dependencies through both forward and backward Mamba encoders, achieving high prediction accuracy. It also achieves competitive or superior full-shot performance compared to task-specific prediction models.
  arXiv  Detail & Related papers  (2024-11-05T09:34:05Z)
• GIFT-Eval: A Benchmark For General Time Series Forecasting Model Evaluation [90.53485251837235]
  GIFT-Eval is a pioneering benchmark aimed at promoting evaluation across diverse datasets. GIFT-Eval encompasses 28 datasets over 144,000 time series and 177 million data points. We also provide a non-leaking pretraining dataset containing approximately 230 billion data points.
  arXiv  Detail & Related papers  (2024-10-14T11:29:38Z)
• A Two-Phase Recall-and-Select Framework for Fast Model Selection [13.385915962994806]
  We propose a two-phase (coarse-recall and fine-selection) model selection framework. It aims to enhance the efficiency of selecting a robust model by leveraging the models' training performances on benchmark datasets. It has been demonstrated that the proposed methodology facilitates the selection of a high-performing model at a rate about 3x times faster than conventional baseline methods.
  arXiv  Detail & Related papers  (2024-03-28T14:44:44Z)
• Robust Fine-Tuning of Vision-Language Models for Domain Generalization [6.7181844004432385]
  Foundation models have impressive zero-shot inference capabilities and robustness under distribution shifts. We present a new recipe for few-shot fine-tuning of the popular vision-language foundation model CLIP. Our experimentation demonstrates that, while zero-shot CLIP fails to match performance of trained vision models on more complex benchmarks, few-shot CLIP fine-tuning outperforms its vision-only counterparts.
  arXiv  Detail & Related papers  (2023-11-03T20:50:40Z)
• Pushing the Limits of Pre-training for Time Series Forecasting in the CloudOps Domain [54.67888148566323]
  We introduce three large-scale time series forecasting datasets from the cloud operations domain. We show it is a strong zero-shot baseline and benefits from further scaling, both in model and dataset size. Accompanying these datasets and results is a suite of comprehensive benchmark results comparing classical and deep learning baselines to our pre-trained method.
  arXiv  Detail & Related papers  (2023-10-08T08:09:51Z)
• AD-PT: Autonomous Driving Pre-Training with Large-scale Point Cloud Dataset [25.935496432142976]
  It is a long-term vision for Autonomous Driving (AD) community that the perception models can learn from a large-scale point cloud dataset. We formulate the point-cloud pre-training task as a semi-supervised problem, which leverages the few-shot labeled and massive unlabeled point-cloud data. We achieve significant performance gains on a series of downstream perception benchmarks including nuScenes, and KITTI, under different baseline models.
  arXiv  Detail & Related papers  (2023-06-01T12:32:52Z)
• MSeg: A Composite Dataset for Multi-domain Semantic Segmentation [100.17755160696939]
  We present MSeg, a composite dataset that unifies semantic segmentation datasets from different domains. We reconcile the generalization and bring the pixel-level annotations into alignment by relabeling more than 220,000 object masks in more than 80,000 images. A model trained on MSeg ranks first on the WildDash-v1 leaderboard for robust semantic segmentation, with no exposure to WildDash data during training.
  arXiv  Detail & Related papers  (2021-12-27T16:16:35Z)
• How Well Do Sparse Imagenet Models Transfer? [75.98123173154605]
  Transfer learning is a classic paradigm by which models pretrained on large "upstream" datasets are adapted to yield good results on "downstream" datasets. In this work, we perform an in-depth investigation of this phenomenon in the context of convolutional neural networks (CNNs) trained on the ImageNet dataset. We show that sparse models can match or even outperform the transfer performance of dense models, even at high sparsities.
  arXiv  Detail & Related papers  (2021-11-26T11:58:51Z)
• Regularizing Generative Adversarial Networks under Limited Data [88.57330330305535]
  This work proposes a regularization approach for training robust GAN models on limited data. We show a connection between the regularized loss and an f-divergence called LeCam-divergence, which we find is more robust under limited training data.
  arXiv  Detail & Related papers  (2021-04-07T17:59:06Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.