{"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/text-encoders-are-performance-bottlenecks-in","title":"Text encoders bottleneck compositionality in contrastive vision-language models","arxiv_id":"2305.14897","date":"2023-05-24","proceeding":null,"authors":["Amita Kamath","Jack Hessel","Kai-Wei Chang"],"abstract":"Performant vision-language (VL) models like CLIP represent captions using a single vector. How much information about language is lost in this bottleneck? We first curate CompPrompts, a set of increasingly compositional image captions that VL models should be able to capture (e.g., single object, to object+property, to multiple interacting objects). Then, we train text-only recovery probes that aim to reconstruct captions from single-vector text representations produced by several VL models. This approach does not require images, allowing us to test on a broader range of scenes compared to prior work. We find that: 1) CLIP's text encoder falls short on more compositional inputs, including object relationships, attribute-object association, counting, and negations; 2) some text encoders work significantly better than others; and 3) text-only recovery performance predicts multi-modal matching performance on ControlledImCaps: a new evaluation benchmark we collect and release consisting of fine-grained compositional images and captions. Specifically, our results suggest text-only recoverability is a necessary (but not sufficient) condition for modeling compositional factors in contrastive VL models. We release our datasets and code.","url_abs":"https://arxiv.org/abs/2305.14897v2","url_pdf":"https://arxiv.org/pdf/2305.14897v2.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":"text-encoders-are-performance-bottlenecks-in","repo_url":"https://github.com/amitakamath/vl_text_encoders_are_bottlenecks","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"attribute","task_name":"Attribute"},{"task_slug":"image-captioning","task_name":"Image Captioning"},{"task_slug":"object","task_name":"Object"}],"methods":[{"method_slug":"clip","method_name":"CLIP"},{"method_slug":"test","method_name":"Test"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2305.14897","atlas_url":"https://app.syntology.ai/?focus=2305.14897","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}