{"id":4284,"date":"2024-08-10T11:35:41","date_gmt":"2024-08-10T03:35:41","guid":{"rendered":"https:\/\/nullthought.net\/?p=4284"},"modified":"2024-12-25T14:30:36","modified_gmt":"2024-12-25T06:30:36","slug":"%e9%80%9a%e8%bf%87%e6%95%b0%e6%8d%ae%e5%a2%9e%e5%bc%ba%e5%af%b9%e6%af%94%e5%be%ae%e8%b0%83%e7%bc%93%e8%a7%a3%e5%af%b9%e8%b1%a1%e5%b9%bb%e8%a7%89%ef%bc%88object-hallucination%ef%bc%89","status":"publish","type":"post","link":"https:\/\/nullthought.net\/?p=4284","title":{"rendered":"\u901a\u8fc7\u6570\u636e\u589e\u5f3a\u5bf9\u6bd4\u5fae\u8c03\u7f13\u89e3\u201c\u5bf9\u8c61\u5e7b\u89c9\u201d\uff08object hallucination\uff09"},"content":{"rendered":"\n<p>\u8bba\u6587<strong><em><a href=\"https:\/\/arxiv.org\/abs\/2405.18654\" target=\"_blank\" rel=\"noreferrer noopener\">Mitigating Object Hallucination via Data Augmented Contrastive Tuning<\/a><\/em><\/strong>\u63a2\u8ba8\u4e86\u5728\u591a\u6a21\u6001\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff08MLLMs\uff09\u4e2d\u5bf9\u8c61\u5e7b\u89c9\u7684\u6311\u6218\u3002\u5bf9\u8c61\u5e7b\u89c9\u662f\u6307\u5f53MLLMs\u751f\u6210\u7684\u63cf\u8ff0\u4e2d\u5305\u62ec\u5e76\u4e0d\u5b58\u5728\u4e8e\u8f93\u5165\u4e2d\u7684\u5bf9\u8c61\u6216\u5176\u5c5e\u6027\uff0c\u8fd9\u5927\u5927\u524a\u5f31\u4e86\u8fd9\u4e9b\u6a21\u578b\u5728\u89c6\u89c9-\u8bed\u8a00\u4efb\u52a1\u4e2d\u7684\u53ef\u9760\u6027\u3002<\/p>\n\n\n\n<p>\u8bba\u6587\u4f5c\u8005\u4e3aPritam Sarkar, Sayna Ebrahimi, Ali Etemad, Ahmad Beirami, Sercan \u00d6. Ar\u0131k, Tomas Pfister\uff0c\u6765\u81eaQueen\u2019s University\uff0cVector Institute\uff0cGoogle Cloud AI Research\u548cGoogle DeepMind\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"360\" src=\"https:\/\/nullthought.net\/wp-content\/uploads\/2024\/08\/image-6-1024x360.png\" alt=\"\" class=\"wp-image-4285\" srcset=\"https:\/\/nullthought.net\/wp-content\/uploads\/2024\/08\/image-6-1024x360.png 1024w, https:\/\/nullthought.net\/wp-content\/uploads\/2024\/08\/image-6-300x106.png 300w, https:\/\/nullthought.net\/wp-content\/uploads\/2024\/08\/image-6-768x270.png 768w, https:\/\/nullthought.net\/wp-content\/uploads\/2024\/08\/image-6.png 1214w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><strong><em><a href=\"https:\/\/arxiv.org\/abs\/2405.18654\" target=\"_blank\" rel=\"noreferrer noopener\">Mitigating Object Hallucination via Data Augmented Contrastive Tuning<\/a><\/em><\/strong><\/figcaption><\/figure>\n\n\n\n<h5 class=\"wp-block-heading\">\u8bba\u6587\u8981\u70b9\u5982\u4e0b\uff1a<\/h5>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u95ee\u9898\u8bc6\u522b<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li>\u5c3d\u7ba1MLLMs\u53d6\u5f97\u4e86\u663e\u8457\u8fdb\u5c55\uff0c\u4f46\u5728\u751f\u6210\u56fe\u50cf\u63cf\u8ff0\u65f6\uff0c\u5b83\u4eec\u5f80\u5f80\u4f1a\u4ea7\u751f\u5e7b\u89c9\u3002\u8fd9\u79cd\u5e7b\u89c9\u901a\u5e38\u6e90\u4e8e\u6a21\u578b\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u5b66\u5230\u7684\u865a\u5047\u5173\u8054\uff0c\u4f8b\u5982\u67d0\u4e9b\u5bf9\u8c61\u7684\u9891\u7e41\u5171\u73b0\uff08\u5982\u201c\u9886\u5e26\u201d\u548c\u201c\u5a5a\u793c\u86cb\u7cd5\u201d\uff09\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>\u4ee5\u5f80\u65b9\u6cd5<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li>\u4ee5\u5f80\u89e3\u51b3\u5e7b\u89c9\u95ee\u9898\u7684\u65b9\u6cd5\u4e3b\u8981\u96c6\u4e2d\u5728\u63a8\u7406\u9636\u6bb5\u3001\u9884\u8bad\u7ec3\u9636\u6bb5\u548c\u5fae\u8c03\u9636\u6bb5\u3002\u7136\u800c\uff0c\u8fd9\u4e9b\u65b9\u6cd5\u901a\u5e38\u4f1a\u589e\u52a0\u63a8\u7406\u65f6\u95f4\u3001\u9700\u8981\u5927\u91cf\u7684\u518d\u8bad\u7ec3\uff0c\u6216\u662f\u4f1a\u964d\u4f4e\u6a21\u578b\u5728\u4e00\u822c\u89c6\u89c9-\u8bed\u8a00\u4efb\u52a1\u4e0a\u7684\u6027\u80fd\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>\u63d0\u51fa\u7684\u89e3\u51b3\u65b9\u6848<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li>\u4f5c\u8005\u63d0\u51fa\u4e86\u4e00\u79cd<strong>\u5bf9\u6bd4\u5fae\u8c03\u65b9\u6cd5<\/strong>\uff0c\u53ef\u5e94\u7528\u4e8e\u9884\u8bad\u7ec3\u7684MLLM\u3002\u6838\u5fc3\u601d\u60f3\u662f\u901a\u8fc7\u751f\u6210\u5f0f\u6570\u636e\u589e\u5f3a\u521b\u5efa\u201c\u5e7b\u89c9\u8bcd\u5143\u201d\uff0c\u9009\u62e9\u6027\u5730\u6539\u53d8\u771f\u5b9e\u4fe1\u606f\uff0c\u7136\u540e\u5fae\u8c03\u6a21\u578b\uff0c\u4f7f\u5176\u66f4\u503e\u5411\u4e8e\u9009\u62e9\u771f\u5b9e\u7684\u8bcd\u5143\u800c\u975e\u5e7b\u89c9\u8bcd\u5143\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>\u65b9\u6cd5\u8bba<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li><strong>\u751f\u6210\u5f0f\u6570\u636e\u589e\u5f3a<\/strong>\uff1a\u901a\u8fc7\u6539\u53d8\u6b63\u786e\u54cd\u5e94\u4e2d\u7684\u5bf9\u8c61\u76f8\u5173\u5c5e\u6027\uff0c\u4f5c\u8005\u751f\u6210\u4e86\u5e7b\u89c9\u54cd\u5e94\uff0c\u4ee5\u521b\u5efa\u5bf9\u6bd4\u5bf9\u3002<\/li>\n\n\n\n<li><strong>\u5bf9\u6bd4\u5fae\u8c03<\/strong>\uff1a\u901a\u8fc7\u5728\u8bcd\u5143\u7ea7\u522b\u5e94\u7528\u5bf9\u6bd4\u635f\u5931\u6765\u5fae\u8c03\u6a21\u578b\uff0c\u51cf\u5c11\u5e7b\u89c9\u8bcd\u5143\u7684\u751f\u6210\u6982\u7387\uff0c\u540c\u65f6\u901a\u8fc7KL\u6563\u5ea6\u6b63\u5219\u5316\u4fdd\u6301\u6a21\u578b\u5728\u89c6\u89c9-\u8bed\u8a00\u4efb\u52a1\u4e2d\u7684\u539f\u59cb\u6027\u80fd\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>\u8bc4\u4f30<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li>\u8be5\u65b9\u6cd5\u5728\u591a\u4e2a\u57fa\u51c6\u6d4b\u8bd5\u4e0a\u8fdb\u884c\u4e86\u9a8c\u8bc1\uff0c\u5305\u62ecCHAIR\u3001MME-Hall\u548cAMBER\uff0c\u7ed3\u679c\u663e\u793a\uff0c\u5728\u4fdd\u6301\u6216\u63d0\u9ad8\u89c6\u89c9-\u8bed\u8a00\u4efb\u52a1\u6027\u80fd\u7684\u540c\u65f6\uff0c\u8be5\u65b9\u6cd5\u663e\u8457\u964d\u4f4e\u4e86\u5e7b\u89c9\u7387\u3002<\/li>\n\n\n\n<li>\u4e0e\u73b0\u6709\u7684HA-DPO\u548cEOS\u65b9\u6cd5\u76f8\u6bd4\uff0c\u63d0\u51fa\u7684\u65b9\u6cd5\uff08\u79f0\u4e3aHALVA\uff09\u5728\u51cf\u8f7b\u5e7b\u89c9\u65b9\u9762\u8868\u73b0\u66f4\u597d\uff0c\u5e76\u4e14\u6ca1\u6709\u964d\u4f4e\u901a\u7528\u4efb\u52a1\u7684\u6027\u80fd\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>\u7ed3\u679c<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li>\u7ed3\u679c\u8868\u660e\uff0cHALVA\u964d\u4f4e\u4e86\u53e5\u5b50\u7ea7\u522b\u7684\u5e7b\u89c9\u7387\uff0c\u5e76\u63d0\u9ad8\u4e86\u56fe\u50cf\u63cf\u8ff0\u4e2d\u5bf9\u8c61\u7684\u8986\u76d6\u7387\u3002<\/li>\n\n\n\n<li>\u5728\u8bf8\u5982\u5b58\u5728\u3001\u5c5e\u6027\u548c\u5173\u7cfb\u8bc6\u522b\u7b49\u8fa8\u522b\u6027\u4efb\u52a1\u4e0a\uff0c\u5b83\u4e5f\u8868\u73b0\u5f97\u66f4\u597d\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>\u6d88\u878d\u7814\u7a76<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li>\u8bba\u6587\u5305\u62ec\u4e86\u5bf9\u65b9\u6cd5\u4e2d\u4e0d\u540c\u7ec4\u4ef6\u7684\u5f71\u54cd\u7684\u6d88\u878d\u7814\u7a76\uff0c\u4f8b\u5982\u5bf9\u6bd4\u635f\u5931\u7684\u9009\u62e9\u548cKL\u6563\u5ea6\u7ea6\u675f\u7684\u5f3a\u5ea6\uff0c\u7814\u7a76\u53d1\u73b0\uff0c\u5e73\u8861\u7684\u65b9\u6cd5\u5bf9\u4e8e\u4fdd\u6301\u6027\u80fd\u81f3\u5173\u91cd\u8981\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>\u66f4\u5e7f\u6cdb\u7684\u5f71\u54cd\u548c\u672a\u6765\u5de5\u4f5c<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li>\u4f5c\u8005\u627f\u8ba4\uff0c\u5c3d\u7ba1\u4ed6\u4eec\u7684\u65b9\u6cd5\u6709\u6548\u5730\u89e3\u51b3\u4e86\u5bf9\u8c61\u5e7b\u89c9\u95ee\u9898\uff0cMLLMs\u4e2d\u7684\u5176\u4ed6\u7c7b\u578b\u7684\u5e7b\u89c9\u4ecd\u7136\u662f\u4e00\u4e2a\u6311\u6218\u3002\u4ed6\u4eec\u5efa\u8bae\uff0c\u672a\u6765\u7684\u5de5\u4f5c\u53ef\u4ee5\u6269\u5c55\u4ed6\u4eec\u7684\u65b9\u6cd5\uff0c\u4ee5\u7f13\u89e3\u4e0d\u540c\u7c7b\u578b\u7684\u5e7b\u89c9\uff0c\u5e76\u63a2\u7d22\u8d85\u8d8aMLLMs\u7684\u5e94\u7528\uff0c\u4f8b\u5982\u51cf\u5c11\u8bed\u8a00\u6a21\u578b\u4e2d\u7684\u504f\u89c1\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"<p>\u8bba\u6587Mitigating Object Hallucination via Data Augmented Co [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","ast-disable-related-posts":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"set","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[8,36],"tags":[39,90,86,80,66],"class_list":["post-4284","post","type-post","status-publish","format-standard","hentry","category-tech","category-36","tag-ai","tag-multimodal","tag-llm","tag-cv","tag-google"],"rttpg_featured_image_url":null,"rttpg_author":{"display_name":"NullThought","author_link":"https:\/\/nullthought.net\/?author=1"},"rttpg_comment":0,"rttpg_category":"<a href=\"https:\/\/nullthought.net\/?cat=8\" rel=\"category\">Tech<\/a> <a href=\"https:\/\/nullthought.net\/?cat=36\" rel=\"category\">\u79d1\u5b66<\/a>","rttpg_excerpt":"\u8bba\u6587Mitigating Object Hallucination via Data Augmented Co&hellip;","_links":{"self":[{"href":"https:\/\/nullthought.net\/index.php?rest_route=\/wp\/v2\/posts\/4284","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/nullthought.net\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/nullthought.net\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/nullthought.net\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/nullthought.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=4284"}],"version-history":[{"count":1,"href":"https:\/\/nullthought.net\/index.php?rest_route=\/wp\/v2\/posts\/4284\/revisions"}],"predecessor-version":[{"id":4286,"href":"https:\/\/nullthought.net\/index.php?rest_route=\/wp\/v2\/posts\/4284\/revisions\/4286"}],"wp:attachment":[{"href":"https:\/\/nullthought.net\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4284"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/nullthought.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4284"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/nullthought.net\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4284"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}