{"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/the-hard-positive-truth-about-vision-language","title":"The Hard Positive Truth about Vision-Language Compositionality","arxiv_id":"2409.17958","date":"2024-09-26","proceeding":null,"authors":["Amita Kamath","Cheng-Yu Hsieh","Kai-Wei Chang","Ranjay Krishna"],"abstract":"Several benchmarks have concluded that our best vision-language models (e.g., CLIP) are lacking in compositionality. Given an image, these benchmarks probe a model's ability to identify its associated caption amongst a set of compositional distractors. In response, a surge of recent proposals show improvements by finetuning CLIP with distractors as hard negatives. Our investigations reveal that these improvements have, in fact, been significantly overstated -- because existing benchmarks do not probe whether finetuned vision-language models remain invariant to hard positives. By curating an evaluation dataset with 112,382 hard negatives and hard positives, we uncover that including hard positives decreases CLIP's performance by 12.9%, while humans perform effortlessly at 99%. CLIP finetuned with hard negatives results in an even larger decrease, up to 38.7%. With this finding, we then produce a 1,775,259 image-text training set with both hard negative and hard positive captions. By training with both, we see improvements on existing benchmarks while simultaneously improving performance on hard positives, indicating a more robust improvement in compositionality. Our work suggests the need for future research to rigorously test and improve CLIP's understanding of semantic relationships between related \"positive\" concepts.","url_abs":"https://arxiv.org/abs/2409.17958v1","url_pdf":"https://arxiv.org/pdf/2409.17958v1.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":"the-hard-positive-truth-about-vision-language","repo_url":"https://github.com/amitakamath/hard_positives","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[],"methods":[{"method_slug":"clip","method_name":"CLIP"},{"method_slug":"set","method_name":"SET"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=2409.17958","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2409.17958"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"deterministic:regex_extraction","url":"https://github.com/amitakamath/hard_positives","reach":{"status":"ok"}}],"summary":{"ran":3,"unverified":1},"by_repo_kind":{"official":{"samples":4,"ran":3,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":4,"samples":[{"code_sha256_prefix":"d52f072444282bb2","entry":"get_image_perturb_fn","repo":"amitakamath/hard_positives","repo_kind":"official","path":"dataset_zoo/perturbations.py","file_url":"https://github.com/amitakamath/hard_positives/blob/HEAD/dataset_zoo/perturbations.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"d52f072444282bb2"}},{"code_sha256_prefix":"0d760ca65749dc14","entry":"pre_caption","repo":"amitakamath/hard_positives","repo_kind":"official","path":"dataset_zoo/retrieval.py","file_url":"https://github.com/amitakamath/hard_positives/blob/HEAD/dataset_zoo/retrieval.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"0d760ca65749dc14"}},{"code_sha256_prefix":"29d83052b338a1f7","entry":"shuffle_rows","repo":"amitakamath/hard_positives","repo_kind":"official","path":"dataset_zoo/perturbations.py","file_url":"https://github.com/amitakamath/hard_positives/blob/HEAD/dataset_zoo/perturbations.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"29d83052b338a1f7"}},{"code_sha256_prefix":"5c8ae0e86a1836f6","entry":"get_text_perturb_fn","repo":"amitakamath/hard_positives","repo_kind":"official","path":"dataset_zoo/perturbations.py","file_url":"https://github.com/amitakamath/hard_positives/blob/HEAD/dataset_zoo/perturbations.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"5c8ae0e86a1836f6"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}