{
  "source_file": "0930",
  "source_platform": "研究机构",
  "author": "Meta AI Research",
  "title_listed": "BART: Denoising Sequence-to-Sequence Pre-training",
  "keywords": "BART,seq2seq,预训练",
  "domain": "文本处理",
  "content_type": "深度技术解析",
  "is_paid_flag": "否",
  "url": "https://ai.meta.com/research/publications/bart-denoising-sequence-to-sequence-pre-training-for-natural-language-generation-translation-and-comprehension/",
  "publish_date": "2019-10",
  "corpus_stage": "预训练/SFT",
  "notes": "BART去噪seq2seq预训练",
  "id": "4df0f718cf4d",
  "collected_at": "2026-10-02T15:54:22",
  "title": "BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension | Facebook AI Research",
  "body_markdown": "Products\nAI Research\nResources\nAbout\nAI Developers\nTry Muse\nResearch\nNLP\nBART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension\nJuly 8, 2020\nAbstract\nWe present BART, a denoising autoencoder for pretraining sequence-to-sequence models. BART is trained by (1) corrupting text with an arbitrary noising function, and (2) learning a model to reconstruct the original text. It uses a standard Tranformer-based neural machine translation architecture which, despite its simplicity, can be seen as generalizing BERT (due to the bidirectional encoder), GPT (with the left-to-right decoder), and other recent pre- training schemes. We evaluate a number of noising approaches, finding the best perfor- mance by both randomly shuffling the order of sentences and using a novel in-filling scheme,where spans of text are replaced with a single mask token. BART is particularly effective when fine tuned for text generation but also works well for comprehension tasks. It matches the performance of RoBERTa on GLUE and SQuAD, and achieves new state-of-the-art results on a range of abstractive di-alogue, question answering, and summarization tasks, with gains of up to 3.5 ROUGE. BART also provides a 1.1 BLEU increase over a back-translation system for machine translation, with only target language pretraining. We also replicate other pretraining schemes within the BART framework, to understand their effect on end-task performance.\nDownload the Paper\nAUTHORS\nWritten by\nMike Lewis\nYinhan Liu\nNaman Goyal\nMarjan Ghazvininejad\nAbdelrahman Mohamed\nOmer Levy\nVes Stoyanov\nLuke Zettlemoyer\nPublisher\nAssociation for Computational Linguistics (ACL)\nResearch Topics\nNatural Language Processing\nRelated Publications\nSeptember 24, 2026\nReinforcement Learni9ng\nMaD-RL: Matching Distributions for Calibrating LLMs with Reinforcement Learning\nSourabh Kulkarni, Ksheeraj Sai Vepuri, Basar Demir, Jason Bohrer, Emily Shen, Jianfa Chen, Nan Jiang, Ankit Jain, Harihar Subramanyam, Mannat Singh, Chirag Nagpal\nSeptember 24, 2026\nRead the Paper\nJuly 17, 2026\nConversational AI\nReinforcement Learni9ng\nLearning to Reason by Analogy via Retrieval-Augmented Reinforcement Fine-Tuning\nZilin Xiao, Qi Ma, Jason Chen, Xintao Chen, Avinash Atreya, Hanjie Chen, Vicente Ordonez\nJuly 17, 2026\nRead the Paper\nJune 05, 2026\nConversational AI\nRanking & Recommendations\nSuperintelligent Retrieval Agent: The Next Frontier of Agentic Retrieval\nZeyu Yang, Qi Ma, Jason Chen, Anshumali Shrivastava\nJune 05, 2026\nRead the Paper\nMay 19, 2026\nHuman & Machine Intelligence\nEgoBabyVLM: Benchmarking Cross-Modal Learning from Naturalistic Egocentric Video Data\nDongyan Lin, Phillip Rust, Angel Villar Corrales, Alvin W. M. Tan, Mahi Luthra, Charles-Eric Saint-James, Rashel Moritz, Sheila Krogh-Jespersen, Vanessa Stark, Surya Parimi, Jiayi Shen, Youssef Benchekroun, Yosuke Higuchi, Martin Gleize, Tom Fizycki, Nicolas Hamilakis, Manel Khentout, Sho Tsuji, Balázs Kégl, Juan Pino, Michael C. Frank, Emmanuel Dupoux\nMay 19, 2026\nRead the Paper\nOctober 31, 2019\nNLP\nFacebook AI's WAT19 Myanmar-English Translation Task Submission\nPeng-Jen Chen, Jiajun Shen,\nMatt Le\n, Vishrav Chaudhary, Ahmed El-Kishky, Guillaume Wenzek,\nMyle Ott\n,\nMarc’Aurelio Ranzato\nOctober 31, 2019\nRead the Paper\nMarch 14, 2019\nNLP\nOn the Pitfalls of Measuring Emergent Communication | Facebook AI Research\nRyan Lowe, Jakob Foerster, Y-Lan Boureau, Joelle Pineau, Yann Dauphin\nMarch 14, 2019\nRead the Paper\nJanuary 13, 2020\nNLP\nScaling up online speech recognition using ConvNets | Facebook AI Research\nVineel Pratap,\nQiantong Xu\n, Jacob Kahn, Gilad Avidov,\nTatiana Likhomanenko\n,\nAwni Hannun\n,\nVitaliy Liptchinsky\n, Gabriel Synnaeve, Ronan Collobert\nJanuary 13, 2020\nRead the Paper\nApril 30, 2018\nNLP\nComputer Vision\nMastering the Dungeon: Grounded Language Learning by Mechanical Turker Descent | Facebook AI Research\nZhilin Yang, Saizheng Zhang, Jack Urbanek, Will Feng,\nAlexander H. Miller\n, Arthur Szlam,\nDouwe Kiela\n,\nJason Weston\nApril 30, 2018\nRead the Paper\nSee All Papers\nHelp Us Pioneer The Future of AI\nWe share our open source frameworks, tools, libraries, and models for everything from research exploration to large-scale production deployment.\nJoin our Team\nMeta AI\nAssistant\nMedia Generation\nVibes\nMuse\nAgent\nAI agents explained\nWhat is agentic AI\nAgentic AI examples\nAI Research\nOverview\nProjects\nResources & tools\nPublications\nGitHub\nResources\nBlog\nLearning Hub\nDemos\nAbout\nOverview\nOpen Source\nCareers\nMeta AI\nMeta AI\nAssistant\nMedia Generation\nVibes\nMuse\nMuse\nAgent\nAI agents explained\nWhat is agentic AI\nAgentic AI examples\nAI Research\nAI Research\nOverview\nProjects\nResources & tools\nPublications\nGitHub\nResources\nBlog\nLearning Hub\nDemos\nAbout\nAbout\nOverview\nOpen Source\nCareers\nPrivacy Policy\nTerms\nCookies\nMeta © 2026",
  "metadata": {
    "title": "BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension | Facebook AI Research",
    "description": "We present BART, a denoising autoencoder for pretraining sequence-to-sequence models. BART is trained by (1) corrupting text with an arbitrary noising...",
    "keywords": null,
    "author": null,
    "og:image": "https://static.xx.fbcdn.net/rsrc.php/v1/yV/r/ljrYCDHmTun.jpg",
    "twitter:card": "summary"
  },
  "paywall": {
    "listed_paid": false,
    "detected": false,
    "marker": "",
    "public_only": false,
    "policy": "付费/会员内容只保留页面公开可见部分，不绕过付费墙"
  },
  "crawl": {
    "success": true,
    "status_code": 200,
    "error": ""
  },
  "code_blocks": [],
  "images": [
    {
      "status": "ok",
      "url": "https://scontent-dfw6-2.xx.fbcdn.net/v/t39.8562-6/252294889_575082167077436_6034106545912333281_n.svg/meta-logo-primary_standardsize.svg?_nc_cat=108&ccb=1-7&_nc_sid=e280be&_nc_ohc=hAHazKbORSwQ7kNvwFs3s3h&_nc_oc=AdpT8oey1lBbgzpbV7ZyVqYXvdIcHIx66czv3z4_y4D5VbmT8-LcxGwGhf3nCEqCFAg&_nc_zt=14&_nc_ht=scontent-dfw6-2.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQPZ9OhQMfPv9AGrsjk0kqrnkmiRVDz1JTrx_oJBwTv-zw&oe=6AD530B9",
      "bytes": 13990,
      "reused": false,
      "local_path": "attachments/4df0f718cf4d/images/body-7e99b2af55f9a99f6f04.svg",
      "src": "https://scontent-dfw6-2.xx.fbcdn.net/v/t39.8562-6/252294889_575082167077436_6034106545912333281_n.svg/meta-logo-primary_standardsize.svg?_nc_cat=108&ccb=1-7&_nc_sid=e280be&_nc_ohc=hAHazKbORSwQ7kNvwFs3s3h&_nc_oc=AdpT8oey1lBbgzpbV7ZyVqYXvdIcHIx66czv3z4_y4D5VbmT8-LcxGwGhf3nCEqCFAg&_nc_zt=14&_nc_ht=scontent-dfw6-2.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQPZ9OhQMfPv9AGrsjk0kqrnkmiRVDz1JTrx_oJBwTv-zw&oe=6AD530B9",
      "alt": "Meta",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://scontent-dfw6-1.xx.fbcdn.net/v/t39.2365-6/91935997_2545519405719235_7197863690534649856_n.jpg?_nc_cat=103&ccb=1-7&_nc_sid=e280be&_nc_ohc=-u0uKb0AgbIQ7kNvwEl5Omk&_nc_oc=AdpToCa5HBs4i1aYEmOTxgjMHtSrjAhrVqvTLDuVy0R1unHhGTu5WBQ4lsaNuq5BloY&_nc_zt=14&_nc_ht=scontent-dfw6-1.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQOV4Tw1cRKRRxUJc9H5oomI0SV9_WYCUWfbVvCwvHV7RA&oe=6AE99EC1",
      "bytes": 288508,
      "reused": false,
      "local_path": "attachments/4df0f718cf4d/images/body-d8e03af299268d6a26da.jpg",
      "src": "https://scontent-dfw6-1.xx.fbcdn.net/v/t39.2365-6/91935997_2545519405719235_7197863690534649856_n.jpg?_nc_cat=103&ccb=1-7&_nc_sid=e280be&_nc_ohc=-u0uKb0AgbIQ7kNvwEl5Omk&_nc_oc=AdpToCa5HBs4i1aYEmOTxgjMHtSrjAhrVqvTLDuVy0R1unHhGTu5WBQ4lsaNuq5BloY&_nc_zt=14&_nc_ht=scontent-dfw6-1.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQOV4Tw1cRKRRxUJc9H5oomI0SV9_WYCUWfbVvCwvHV7RA&oe=6AE99EC1",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://scontent-dfw6-1.xx.fbcdn.net/v/t39.2365-6/87524316_2677189655726266_6338721200264445952_n.svg?_nc_cat=103&ccb=1-7&_nc_sid=e280be&_nc_ohc=fVw4LJojBokQ7kNvwGoeBzo&_nc_oc=AdrkP6I-04u6p4Jook2sPg73GEI1IfiIRn2GycI6WgIkObCoxv9tkmBFc09yToCgBiA&_nc_zt=14&_nc_ht=scontent-dfw6-1.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQMpvEnZPPAdVbip6Vx_v4ZlQK7La8NNEiiV4sZDpDQRfA&oe=6AE99A78",
      "bytes": 768,
      "reused": false,
      "local_path": "attachments/4df0f718cf4d/images/body-301c827e723c3b33d373.svg",
      "src": "https://scontent-dfw6-1.xx.fbcdn.net/v/t39.2365-6/87524316_2677189655726266_6338721200264445952_n.svg?_nc_cat=103&ccb=1-7&_nc_sid=e280be&_nc_ohc=fVw4LJojBokQ7kNvwGoeBzo&_nc_oc=AdrkP6I-04u6p4Jook2sPg73GEI1IfiIRn2GycI6WgIkObCoxv9tkmBFc09yToCgBiA&_nc_zt=14&_nc_ht=scontent-dfw6-1.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQMpvEnZPPAdVbip6Vx_v4ZlQK7La8NNEiiV4sZDpDQRfA&oe=6AE99A78",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://scontent-dfw5-2.xx.fbcdn.net/v/t39.8562-6/335682312_964107378293184_3093631164486164913_n.svg?_nc_cat=100&ccb=1-7&_nc_sid=e280be&_nc_ohc=fYUd8cfKa3cQ7kNvwHzxyUo&_nc_oc=AdqO5CXvI_Ry9-R7gLxXAricVHfV9jgZI4bhWqxGsfabwKdJ_6gU1IS1JNw7Yk5A8rI&_nc_zt=14&_nc_ht=scontent-dfw5-2.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQNLCyC4oVnn6kyReiHf6_c5jSDZU5H7ZQgF6UJmzZyRwQ&oe=6AD520E7",
      "bytes": 536,
      "reused": false,
      "local_path": "attachments/4df0f718cf4d/images/body-94cacf445f5319830a4b.svg",
      "src": "https://scontent-dfw5-2.xx.fbcdn.net/v/t39.8562-6/335682312_964107378293184_3093631164486164913_n.svg?_nc_cat=100&ccb=1-7&_nc_sid=e280be&_nc_ohc=fYUd8cfKa3cQ7kNvwHzxyUo&_nc_oc=AdqO5CXvI_Ry9-R7gLxXAricVHfV9jgZI4bhWqxGsfabwKdJ_6gU1IS1JNw7Yk5A8rI&_nc_zt=14&_nc_ht=scontent-dfw5-2.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQNLCyC4oVnn6kyReiHf6_c5jSDZU5H7ZQgF6UJmzZyRwQ&oe=6AD520E7",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://scontent-dfw6-1.xx.fbcdn.net/v/t39.8562-6/336009607_1870102080040414_6753977241281150924_n.svg?_nc_cat=103&ccb=1-7&_nc_sid=e280be&_nc_ohc=EOxnus0Vys8Q7kNvwFGrms5&_nc_oc=AdodxvUHQnETZW32TOMmP5S_noQxTy-8rzUluvcvOmKo3rO2-NuuXTQLrgMGVVO-y84&_nc_zt=14&_nc_ht=scontent-dfw6-1.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQP9VQ5OBHZZ2BFJC8P8nCtUwRVtUFpAKhZjOQP4o_AH5Q&oe=6AD51922",
      "bytes": 1126,
      "reused": false,
      "local_path": "attachments/4df0f718cf4d/images/body-7868081baf933d2a0c51.svg",
      "src": "https://scontent-dfw6-1.xx.fbcdn.net/v/t39.8562-6/336009607_1870102080040414_6753977241281150924_n.svg?_nc_cat=103&ccb=1-7&_nc_sid=e280be&_nc_ohc=EOxnus0Vys8Q7kNvwFGrms5&_nc_oc=AdodxvUHQnETZW32TOMmP5S_noQxTy-8rzUluvcvOmKo3rO2-NuuXTQLrgMGVVO-y84&_nc_zt=14&_nc_ht=scontent-dfw6-1.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQP9VQ5OBHZZ2BFJC8P8nCtUwRVtUFpAKhZjOQP4o_AH5Q&oe=6AD51922",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://scontent-dfw6-2.xx.fbcdn.net/v/t39.8562-6/336289415_1541032296405649_2165099305308791297_n.svg?_nc_cat=109&ccb=1-7&_nc_sid=e280be&_nc_ohc=szNSkRClPB8Q7kNvwG_QuzC&_nc_oc=AdrwcmNZ6VBPiUuOKpS80tKMlMpN-566glDgJUkbOeVxfK_YcxcnHGmQXAT2VLrdJy4&_nc_zt=14&_nc_ht=scontent-dfw6-2.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQOfgxS1bDyhckUR3b0vdnl9t-VSHI_vsKwG-kPb6E1bWw&oe=6AD544FB",
      "bytes": 806,
      "reused": false,
      "local_path": "attachments/4df0f718cf4d/images/body-4f50a5f77d2284d4f927.svg",
      "src": "https://scontent-dfw6-2.xx.fbcdn.net/v/t39.8562-6/336289415_1541032296405649_2165099305308791297_n.svg?_nc_cat=109&ccb=1-7&_nc_sid=e280be&_nc_ohc=szNSkRClPB8Q7kNvwG_QuzC&_nc_oc=AdrwcmNZ6VBPiUuOKpS80tKMlMpN-566glDgJUkbOeVxfK_YcxcnHGmQXAT2VLrdJy4&_nc_zt=14&_nc_ht=scontent-dfw6-2.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQOfgxS1bDyhckUR3b0vdnl9t-VSHI_vsKwG-kPb6E1bWw&oe=6AD544FB",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://scontent-dfw6-2.xx.fbcdn.net/v/t39.8562-6/335648731_142576991793348_7786819189843639239_n.svg?_nc_cat=108&ccb=1-7&_nc_sid=e280be&_nc_ohc=kR8jmDz5Tu8Q7kNvwFsKqaM&_nc_oc=AdqrOm2ElI15Ej13sM8SirbTVZiUpR5DMxc4KAJkrz9qH4EpZnouLCCNEJvPdLyTI_o&_nc_zt=14&_nc_ht=scontent-dfw6-2.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQOcloktlVj8H37stZk7VQNhtG6nPh8Co09YBUunLkMgWw&oe=6AD5262E",
      "bytes": 771,
      "reused": false,
      "local_path": "attachments/4df0f718cf4d/images/body-31dd7ad5d9ea0b79b3fc.svg",
      "src": "https://scontent-dfw6-2.xx.fbcdn.net/v/t39.8562-6/335648731_142576991793348_7786819189843639239_n.svg?_nc_cat=108&ccb=1-7&_nc_sid=e280be&_nc_ohc=kR8jmDz5Tu8Q7kNvwFsKqaM&_nc_oc=AdqrOm2ElI15Ej13sM8SirbTVZiUpR5DMxc4KAJkrz9qH4EpZnouLCCNEJvPdLyTI_o&_nc_zt=14&_nc_ht=scontent-dfw6-2.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQOcloktlVj8H37stZk7VQNhtG6nPh8Co09YBUunLkMgWw&oe=6AD5262E",
      "alt": "",
      "downloaded": true
    }
  ],
  "audio": [],
  "videos": [
    {
      "url": "https://www.youtube.com/@aiatmeta",
      "src": "https://www.youtube.com/@aiatmeta",
      "status": "reference_only",
      "kind": "channel_reference"
    }
  ],
  "links": [
    {
      "href": "https://developer.meta.com/ai/?utm_source=ai_meta_site&utm_medium=web&utm_campaign=nav_AI-developers_07242026&utm_content=nav_AI-developers",
      "label": "AI Developers",
      "type": "a"
    },
    {
      "href": "https://applink.muse.ai/?pt=10684&pid=ai_meta_site&utm_source=ai_meta_site&utm_medium=web&utm_campaign=nav_try-muse_09082026&utm_content=nav_try-muse_09082026&ct=nav_try-muse_09082026&referrer=utm_source%3Dai_meta_site%26utm_medium%3Dweb%26utm_campaign%3Dnav_try-muse_09082026%26utm_content%3Dnav_try-muse_09082026",
      "label": "Try Muse",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://scontent-dfw5-1.xx.fbcdn.net/v/t39.8562-6/106373513_3414102562251474_8005430471454563564_n.pdf?_nc_cat=105&ccb=1-7&_nc_sid=b8d81d&_nc_ohc=pDajRUPyUoYQ7kNvwGk2SXV&_nc_oc=AdpjmqgYgzwbjpWfsXWG6ErMCFa_b8gC8PR_EsmRlthvYrzwZmOrOur2lIpRS14RoAw&_nc_zt=14&_nc_ht=scontent-dfw5-1.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQN2wuvfTgMRpWzOFOhZFFcO-_FpaBEBpGBULODGuP2Rrg&oe=6AD529C4",
      "label": "Download the Paper",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/mike-lewis/",
      "label": "Mike Lewis",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/yinhan-liu/",
      "label": "Yinhan Liu",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/marjan-ghazvininejad/",
      "label": "Marjan Ghazvininejad",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/luke-zettlemoyer/",
      "label": "Luke Zettlemoyer",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/research/NLP/",
      "label": "Natural Language Processing",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/research/publications/mad-rl-matching-distributions-for-calibrating-llms-with-reinforcement-learning/",
      "label": "Read the Paper",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/research/publications/learning-to-reason-by-analogy-via-retrieval-augmented-reinforcement-fine-tuning/",
      "label": "Read the Paper",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/research/publications/superintelligent-retrieval-agent-the-next-frontier-of-agentic-retrieval/",
      "label": "Read the Paper",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/research/publications/egobabyvlm-benchmarking-cross-modal-learning-from-naturalistic-egocentric-video-data/",
      "label": "Read the Paper",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/matt-le/",
      "label": "Matt Le",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/myle-ott/",
      "label": "Myle Ott",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/marc-aurelio-ranzato/",
      "label": "Marc’Aurelio Ranzato",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/research/publications/facebook-ai-wat19-myanmar-english-translation-task-submission/",
      "label": "Read the Paper",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/research/publications/on-the-pitfalls-of-measuring-emergent-communication/",
      "label": "Read the Paper",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/qiantong-xu",
      "label": "Qiantong Xu",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/tatiana-likhomanenko",
      "label": "Tatiana Likhomanenko",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/awni-hannun",
      "label": "Awni Hannun",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/vitaliy-liptchinsky",
      "label": "Vitaliy Liptchinsky",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/research/publications/scaling-up-online-speech-recognition-using-convnets/",
      "label": "Read the Paper",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/alexander-h-miller",
      "label": "Alexander H. Miller",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/douwe-kiela",
      "label": "Douwe Kiela",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/people/jason-weston",
      "label": "Jason Weston",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/research/publications/mastering-the-dungeon-grounded-language-learning-by-mechanical-turker-descent/",
      "label": "Read the Paper",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/results/?q&content_types[0]=publication&sort_by=relevance&view=list&page=1",
      "label": "See All Papers",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/join-us/",
      "label": "Join our Team",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/meta-ai/assistant/",
      "label": "Meta AI",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/meta-ai/media-generation/",
      "label": "Media Generation",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/meta-ai/vibes/",
      "label": "Vibes",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/muse/",
      "label": "Muse",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/learn/agentic-ai/what-are-ai-agents/",
      "label": "AI agents explained",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/learn/agentic-ai/what-is-agentic-ai/",
      "label": "What is agentic AI",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/learn/agentic-ai/agentic-ai-examples/",
      "label": "Agentic AI examples",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/research",
      "label": "AI Research",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/research/",
      "label": "Overview",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/resources/",
      "label": "Resources & tools",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/results/?content_types%5B0%5D=publication",
      "label": "Publications",
      "type": "a"
    },
    {
      "href": "https://github.com/facebookresearch/",
      "label": "GitHub",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/blog/",
      "label": "Blog",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/learn/",
      "label": "Learning Hub",
      "type": "a"
    },
    {
      "href": "https://aidemos.meta.com/",
      "label": "Demos",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/about",
      "label": "About",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/about/",
      "label": "Overview",
      "type": "a"
    },
    {
      "href": "https://ai.meta.com/opensourceai/",
      "label": "Open Source",
      "type": "a"
    },
    {
      "href": "https://www.metacareers.com/jobs/?is_leadership=0&sub_teams[0]=Artificial%20Intelligence&is_in_page=0",
      "label": "Careers",
      "type": "a"
    },
    {
      "href": "https://www.facebook.com/aiatmeta/",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://twitter.com/aiatmeta/",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://www.linkedin.com/showcase/aiatmeta",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/@aiatmeta",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://www.facebook.com/about/privacy/",
      "label": "Privacy Policy",
      "type": "a"
    },
    {
      "href": "https://www.facebook.com/policies/",
      "label": "Terms",
      "type": "a"
    },
    {
      "href": "https://www.facebook.com/policies/cookies/",
      "label": "Cookies",
      "type": "a"
    }
  ],
  "repair_version": 2,
  "repaired_at": "2026-10-14T10:53:41Z",
  "source_url": "https://ai.meta.com/research/publications/bart-denoising-sequence-to-sequence-pre-training-for-natural-language-generation-translation-and-comprehension/",
  "papers": [
    {
      "status": "ok",
      "url": "https://scontent-dfw5-1.xx.fbcdn.net/v/t39.8562-6/106373513_3414102562251474_8005430471454563564_n.pdf?_nc_cat=105&ccb=1-7&_nc_sid=b8d81d&_nc_ohc=pDajRUPyUoYQ7kNvwGk2SXV&_nc_oc=AdpjmqgYgzwbjpWfsXWG6ErMCFa_b8gC8PR_EsmRlthvYrzwZmOrOur2lIpRS14RoAw&_nc_zt=14&_nc_ht=scontent-dfw5-1.xx&_nc_gid=D7c6uSFn8kQq8wumDKjupg&_nc_ss=7820f&oh=00_AQN2wuvfTgMRpWzOFOhZFFcO-_FpaBEBpGBULODGuP2Rrg&oe=6AD529C4",
      "bytes": 235354,
      "reused": false,
      "local_path": "papers/repair/8a28f728f14812e8b1f3.pdf"
    }
  ],
  "actual_video_files": [],
  "body_chars": 4826,
  "expected_body_images": 7,
  "discovery_status": "ok",
  "delivery_status": "ok",
  "missing_reason": "",
  "raw_html": "attachments/4df0f718cf4d/pages/source.html",
  "local_html": "attachments/4df0f718cf4d/pages/article.html",
  "body_base_path": "attachments/4df0f718cf4d/pages/",
  "body_markdown_path": "attachments/4df0f718cf4d/pages/article.md"
}