{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/exploring-the-limits-of-out-of-distribution","title":"Exploring the Limits of Out-of-Distribution Detection","arxiv_id":"2106.03004","date":"2021-06-06","proceeding":"NeurIPS 2021 12","authors":["Stanislav Fort","Jie Ren","Balaji Lakshminarayanan"],"abstract":"Near out-of-distribution detection (OOD) is a major challenge for deep neural networks. We demonstrate that large-scale pre-trained transformers can significantly improve the state-of-the-art (SOTA) on a range of near OOD tasks across different data modalities. For instance, on CIFAR-100 vs CIFAR-10 OOD detection, we improve the AUROC from 85% (current SOTA) to more than 96% using Vision Transformers pre-trained on ImageNet-21k. On a challenging genomics OOD detection benchmark, we improve the AUROC from 66% to 77% using transformers and unsupervised pre-training. To further improve performance, we explore the few-shot outlier exposure setting where a few examples from outlier classes may be available; we show that pre-trained transformers are particularly well-suited for outlier exposure, and that the AUROC of OOD detection on CIFAR-100 vs CIFAR-10 can be improved to 98.7% with just 1 image per OOD class, and 99.46% with 10 images per OOD class. For multi-modal image-text pre-trained transformers such as CLIP, we explore a new way of using just the names of outlier classes as a sole source of information without any accompanying images, and show that this outperforms previous SOTA on standard vision OOD benchmark tasks.","url_abs":"https://arxiv.org/abs/2106.03004v3","url_pdf":"https://arxiv.org/pdf/2106.03004v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"exploring-the-limits-of-out-of-distribution","repo_url":"https://github.com/stanislavfort/exploring_the_limits_of_OOD_detection","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"jax","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"ood-detection","task_name":"Out of Distribution (OOD) Detection"},{"task_slug":"out-of-distribution-detection","task_name":"Out-of-Distribution Detection"},{"task_slug":"unsupervised-pre-training","task_name":"Unsupervised Pre-training"}],"methods":[{"method_slug":"clip","method_name":"CLIP"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/out-of-distribution-detection-on-cifar-10-vs","task":"Out-of-Distribution Detection","dataset":"CIFAR-10 vs CIFAR-100","model":"R+ViT finetuned on CIFAR-10","rank_in_archive_order":2,"of":14,"metrics":{"AUPR":"97.75","AUROC":"98.52"},"uses_additional_data":true},{"leaderboard":"/sota/out-of-distribution-detection-on-cifar-10-vs","task":"Out-of-Distribution Detection","dataset":"CIFAR-10 vs CIFAR-100","model":"ViT finetuned on CIFAR-10","rank_in_archive_order":3,"of":14,"metrics":{"AUPR":"97.68","AUROC":"98.42"},"uses_additional_data":true},{"leaderboard":"/sota/out-of-distribution-detection-on-cifar-10-vs","task":"Out-of-Distribution Detection","dataset":"CIFAR-10 vs CIFAR-100","model":"MLP-Mixer finetuned on CIFAR-10","rank_in_archive_order":4,"of":14,"metrics":{"AUPR":"96.28","AUROC":"97.85"},"uses_additional_data":true},{"leaderboard":"/sota/out-of-distribution-detection-on-cifar-100-vs","task":"Out-of-Distribution Detection","dataset":"CIFAR-100 vs CIFAR-10","model":"Ensemble of ViTs","rank_in_archive_order":2,"of":14,"metrics":{"AUROC":"98.11"},"uses_additional_data":true},{"leaderboard":"/sota/out-of-distribution-detection-on-cifar-100-vs","task":"Out-of-Distribution Detection","dataset":"CIFAR-100 vs CIFAR-10","model":"ViT-L_16 finetuned on CIFAR-100","rank_in_archive_order":3,"of":14,"metrics":{"AUROC":"97.98"},"uses_additional_data":true},{"leaderboard":"/sota/out-of-distribution-detection-on-cifar-100-vs","task":"Out-of-Distribution Detection","dataset":"CIFAR-100 vs CIFAR-10","model":"R50+ViT_B-16 finetuned on CIFAR-100","rank_in_archive_order":4,"of":14,"metrics":{"AUPR":"92.08","AUROC":"96.23"},"uses_additional_data":true},{"leaderboard":"/sota/out-of-distribution-detection-on-cifar-100-vs","task":"Out-of-Distribution Detection","dataset":"CIFAR-100 vs CIFAR-10","model":"ViT_B-16 finetuned on CIFAR-100","rank_in_archive_order":5,"of":14,"metrics":{"AUPR":"91.89","AUROC":"95.53"},"uses_additional_data":true},{"leaderboard":"/sota/out-of-distribution-detection-on-cifar-100-vs","task":"Out-of-Distribution Detection","dataset":"CIFAR-100 vs CIFAR-10","model":"MLP-Mixer_B-16 finetuned on CIFAR-100","rank_in_archive_order":6,"of":14,"metrics":{"AUPR":"90.22","AUROC":"95.31"},"uses_additional_data":true},{"leaderboard":"/sota/out-of-distribution-detection-on-cifar-100-vs","task":"Out-of-Distribution Detection","dataset":"CIFAR-100 vs CIFAR-10","model":"CLIP using class name words describing the two distributions","rank_in_archive_order":7,"of":14,"metrics":{"AUROC":"94.68"},"uses_additional_data":true}],"syntology":{"syntology_url":"https://syntology.ai/paper/2106.03004","atlas_url":"https://app.syntology.ai/?focus=2106.03004","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}