Does CLIP's Generalization Performance Mainly Stem from High Train-Test Similarity?

• URL: http://arxiv.org/abs/2310.09562v2
• Date: Thu, 14 Mar 2024 18:18:49 GMT
• Title: Does CLIP's Generalization Performance Mainly Stem from High Train-Test Similarity?
• Authors: Prasanna Mayilvahanan, Thaddäus Wiedemer, Evgenia Rusak, Matthias Bethge, Wieland Brendel,
• Abstract summary: Foundation models like CLIP are trained on hundreds of millions of samples and effortlessly generalize to new tasks and inputs. We show that high train-test similarity is insufficient to explain CLIP's OOD performance.
• Score: 21.556266136890844
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Foundation models like CLIP are trained on hundreds of millions of samples and effortlessly generalize to new tasks and inputs. Out of the box, CLIP shows stellar zero-shot and few-shot capabilities on a wide range of out-of-distribution (OOD) benchmarks, which prior works attribute mainly to today's large and comprehensive training dataset (like LAION). However, it is questionable how meaningful terms like out-of-distribution generalization are for CLIP as it seems likely that web-scale datasets like LAION simply contain many samples that are similar to common OOD benchmarks originally designed for ImageNet. To test this hypothesis, we retrain CLIP on pruned LAION splits that replicate ImageNet's train-test similarity with respect to common OOD benchmarks. While we observe a performance drop on some benchmarks, surprisingly, CLIP's overall performance remains high. This shows that high train-test similarity is insufficient to explain CLIP's OOD performance, and other properties of the training data must drive CLIP to learn more generalizable representations. Additionally, by pruning data points that are dissimilar to the OOD benchmarks, we uncover a 100M split of LAION ($\frac{1}{4}$th of its original size) on which CLIP can be trained to match its original OOD performance.

Related papers

• Do CLIPs Always Generalize Better than ImageNet Models? [45.87070442259975]
  Large vision language models, such as CLIPs, have revolutionized modern machine learning. We show that CLIPs trained on either LAION or the OpenAI data exhibit notable performance drops on the counter group. Surprisingly, we observe that single-modal models trained on ImageNet are more robust than CLIPs.
  arXiv  Detail & Related papers  (2024-03-18T06:04:02Z)
• A Hard-to-Beat Baseline for Training-free CLIP-based Adaptation [121.0693322732454]
  Contrastive Language-Image Pretraining (CLIP) has gained popularity for its remarkable zero-shot capacity. Recent research has focused on developing efficient fine-tuning methods to enhance CLIP's performance in downstream tasks. We revisit a classical algorithm, Gaussian Discriminant Analysis (GDA), and apply it to the downstream classification of CLIP.
  arXiv  Detail & Related papers  (2024-02-06T15:45:27Z)
• Adapting Contrastive Language-Image Pretrained (CLIP) Models for Out-of-Distribution Detection [1.597617022056624]
  We present a comprehensive experimental study on pretrained feature extractors for visual out-of-distribution (OOD) detection. We propose a new simple and scalable method called textitpseudo-label probing (PLP) that adapts vision-language models for OOD detection.
  arXiv  Detail & Related papers  (2023-03-10T10:02:18Z)
• CLIPood: Generalizing CLIP to Out-of-Distributions [73.86353105017076]
  Contrastive language-image pre-training (CLIP) models have shown impressive zero-shot ability, but the further adaptation of CLIP on downstream tasks undesirably degrades OOD performances. We propose CLIPood, a fine-tuning method that can adapt CLIP models to OOD situations where both domain shifts and open classes may occur on unseen test data. Experiments on diverse datasets with different OOD scenarios show that CLIPood consistently outperforms existing generalization techniques.
  arXiv  Detail & Related papers  (2023-02-02T04:27:54Z)
• CLIP Itself is a Strong Fine-tuner: Achieving 85.7% and 88.0% Top-1 Accuracy with ViT-B and ViT-L on ImageNet [139.56863124214905]
  We find that fine-tuning performance of CLIP is substantially underestimated. Specifically, CLIP ViT-Base/16 and CLIP ViT-Large/14 can achieve 85.7%,88.0% finetuning Top-1 accuracy on the ImageNet-1K dataset.
  arXiv  Detail & Related papers  (2022-12-12T18:59:59Z)
• OrdinalCLIP: Learning Rank Prompts for Language-Guided Ordinal Regression [94.28253749970534]
  We propose to learn the rank concepts from the rich semantic CLIP latent space. OrdinalCLIP consists of learnable context tokens and learnable rank embeddings. Experimental results show that our paradigm achieves competitive performance in general ordinal regression tasks.
  arXiv  Detail & Related papers  (2022-06-06T03:54:53Z)
• Democratizing Contrastive Language-Image Pre-training: A CLIP Benchmark of Data, Model, and Supervision [26.13829720290035]
  Contrastive Language-Image Pretraining (CLIP) has emerged as a novel paradigm to learn visual models from language supervision. We propose CLIP-benchmark, a first attempt to evaluate, analyze, and benchmark CLIP and its variants.
  arXiv  Detail & Related papers  (2022-03-11T08:41:00Z)
• The CLEAR Benchmark: Continual LEArning on Real-World Imagery [77.98377088698984]
  Continual learning (CL) is widely regarded as crucial challenge for lifelong AI. We introduce CLEAR, the first continual image classification benchmark dataset with a natural temporal evolution of visual concepts. We find that a simple unsupervised pre-training step can already boost state-of-the-art CL algorithms.
  arXiv  Detail & Related papers  (2022-01-17T09:09:09Z)

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.