{"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/object-hallucination-in-image-captioning","title":"Object Hallucination in Image Captioning","arxiv_id":"1809.02156","date":"2018-09-06","proceeding":"EMNLP 2018 10","authors":["Anna Rohrbach","Lisa Anne Hendricks","Kaylee Burns","Trevor Darrell","Kate Saenko"],"abstract":"Despite continuously improving performance, contemporary image captioning\nmodels are prone to \"hallucinating\" objects that are not actually in a scene.\nOne problem is that standard metrics only measure similarity to ground truth\ncaptions and may not fully capture image relevance. In this work, we propose a\nnew image relevance metric to evaluate current models with veridical visual\nlabels and assess their rate of object hallucination. We analyze how captioning\nmodel architectures and learning objectives contribute to object hallucination,\nexplore when hallucination is likely due to image misclassification or language\npriors, and assess how well current sentence metrics capture object\nhallucination. We investigate these questions on the standard image captioning\nbenchmark, MSCOCO, using a diverse set of models. Our analysis yields several\ninteresting findings, including that models which score best on standard\nsentence metrics do not always have lower hallucination and that models which\nhallucinate more tend to make errors driven by language priors.","url_abs":"http://arxiv.org/abs/1809.02156v2","url_pdf":"http://arxiv.org/pdf/1809.02156v2.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":"object-hallucination-in-image-captioning","repo_url":"https://github.com/LisaAnne/Hallucination","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"hallucination","task_name":"Hallucination"},{"task_slug":"image-captioning","task_name":"Image Captioning"},{"task_slug":"object","task_name":"Object"},{"task_slug":"object-hallucination","task_name":"Object Hallucination"},{"task_slug":"sentence","task_name":"Sentence"}],"methods":[],"datasets_introduced":[{"slug":"object-halbench","name":"Object HalBench","full_name":""}],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1809.02156","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}