Deprecated: The each() function is deprecated. This message will be suppressed on further calls in /home/zhenxiangba/zhenxiangba.com/public_html/phproxy-improved-master/index.php on line 456 Paper page - Aurora-M: The First Open Source Multilingual Language Model Red-teamed
according to the U.S. Executive Order
Please give a thumbs up to this comment if you found it helpful!
\n
If you want recommendations for any Paper on Hugging Face checkout this Space
\n
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: \n\n@librarian-bot\n\t recommend
\n","updatedAt":"2024-04-03T01:22:32.881Z","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":318,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7214733958244324},"editors":["librarian-bot"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg"],"reactions":[{"reaction":"👍","users":["mayank-mishra","Xa9aX","tanmaylaud","sted97","appleleo"],"count":5}],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2404.00399","authors":[{"_id":"660b5b0becc38ac2450f232c","user":{"_id":"6308c49c454dc257521bc7f9","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6308c49c454dc257521bc7f9/UWUS6OPa6OpVu1T0gd-wJ.jpeg","isPro":false,"fullname":"Taishi","user":"Taishi-N324","type":"user"},"name":"Taishi Nakamura","status":"claimed_verified","statusLastChangedAt":"2024-04-05T10:05:04.200Z","hidden":false},{"_id":"660b5b0becc38ac2450f232d","user":{"_id":"62cd5057674cdb524450093d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/62cd5057674cdb524450093d/f67rlrdsKPRTLdXCXoa_X.jpeg","isPro":false,"fullname":"Mayank Mishra","user":"mayank-mishra","type":"user"},"name":"Mayank Mishra","status":"claimed_verified","statusLastChangedAt":"2024-04-02T11:05:31.391Z","hidden":false},{"_id":"660b5b0becc38ac2450f232e","user":{"_id":"61b85aa99ba538c73a7dc78b","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/61b85aa99ba538c73a7dc78b/gWxtQAvOYn7cXgE_nAy0p.jpeg","isPro":false,"fullname":"Simone Tedeschi","user":"sted97","type":"user"},"name":"Simone Tedeschi","status":"claimed_verified","statusLastChangedAt":"2024-04-02T11:07:48.747Z","hidden":false},{"_id":"660b5b0becc38ac2450f232f","user":{"_id":"626a6ce8adf559c88de9ace0","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1651142275247-626a6ce8adf559c88de9ace0.jpeg","isPro":false,"fullname":"CYK","user":"cyk1337","type":"user"},"name":"Yekun Chai","status":"admin_assigned","statusLastChangedAt":"2024-04-02T15:11:18.357Z","hidden":true},{"_id":"660b5b0becc38ac2450f2330","user":{"_id":"6178d842267320abb99df776","avatarUrl":"/avatars/8739705ef7eabc945a1ef3c07196af13.svg","isPro":false,"fullname":"Jason Stillerman","user":"stillerman","type":"user"},"name":"Jason T Stillerman","status":"claimed_verified","statusLastChangedAt":"2024-04-03T09:26:11.333Z","hidden":false},{"_id":"660b5b0becc38ac2450f2331","user":{"_id":"62e7dd4036a8e8a82700041c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/62e7dd4036a8e8a82700041c/Dgk9mXYLVd4LpiNLWjn-q.jpeg","isPro":false,"fullname":"Felix Friedrich","user":"felfri","type":"user"},"name":"Felix Friedrich","status":"claimed_verified","statusLastChangedAt":"2025-06-20T12:21:21.925Z","hidden":false},{"_id":"660b5b0becc38ac2450f2332","name":"Prateek Yadav","hidden":false},{"_id":"660b5b0becc38ac2450f2333","user":{"_id":"5fd85dbada7d9193d8b05b69","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1624471791171-5fd85dbada7d9193d8b05b69.png","isPro":false,"fullname":"Tanmay Laud","user":"tanmaylaud","type":"user"},"name":"Tanmay Laud","status":"claimed_verified","statusLastChangedAt":"2024-04-03T09:26:26.444Z","hidden":false},{"_id":"660b5b0becc38ac2450f2334","user":{"_id":"60535c9d10aba34e3b6a2ef7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1616075850230-noauth.jpeg","isPro":false,"fullname":"vumichien","user":"vumichien","type":"user"},"name":"Vu Minh Chien","status":"claimed_verified","statusLastChangedAt":"2024-04-02T19:35:16.724Z","hidden":false},{"_id":"660b5b0becc38ac2450f2335","user":{"_id":"62b7fb545233925f253531c8","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/62b7fb545233925f253531c8/W50u2G1HK3EtUKHRU189V.jpeg","isPro":false,"fullname":"Terry Yue Zhuo","user":"terryyz","type":"user"},"name":"Terry Yue Zhuo","status":"claimed_verified","statusLastChangedAt":"2025-08-27T07:12:50.441Z","hidden":false},{"_id":"660b5b0becc38ac2450f2336","user":{"_id":"643006f01572f43a481766a9","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/643006f01572f43a481766a9/egLlzmsWxHovmuvqAPnBO.jpeg","isPro":false,"fullname":"_","user":"Xa9aX","type":"user"},"name":"Diganta Misra","status":"extracted_confirmed","statusLastChangedAt":"2024-04-02T01:15:26.036Z","hidden":false},{"_id":"660b5b0becc38ac2450f2337","name":"Ben Bogin","hidden":false},{"_id":"660b5b0becc38ac2450f2338","name":"Xuan-Son Vu","hidden":false},{"_id":"660b5b0becc38ac2450f2339","user":{"_id":"62293b03acd5bef90e55c4ae","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1667593978543-62293b03acd5bef90e55c4ae.png","isPro":false,"fullname":"Marzena Karpinska","user":"marzena","type":"user"},"name":"Marzena Karpinska","status":"claimed_verified","statusLastChangedAt":"2024-07-01T07:47:18.567Z","hidden":false},{"_id":"660b5b0becc38ac2450f233a","user":{"_id":"6303e701d14428368d19569e","avatarUrl":"/avatars/8fb08261884c5fc61925a25594161f13.svg","isPro":false,"fullname":"Arnav Dantuluri","user":"Aldrich","type":"user"},"name":"Arnav Varma Dantuluri","status":"claimed_verified","statusLastChangedAt":"2024-04-08T08:12:24.997Z","hidden":false},{"_id":"660b5b0becc38ac2450f233b","user":{"_id":"611669ad267478fd54891f8d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1659021160775-611669ad267478fd54891f8d.jpeg","isPro":false,"fullname":"Wojciech Kusa","user":"WojciechKusa","type":"user"},"name":"Wojciech Kusa","status":"claimed_verified","statusLastChangedAt":"2025-06-11T09:23:29.694Z","hidden":false},{"_id":"660b5b0becc38ac2450f233c","name":"Tommaso Furlanello","hidden":false},{"_id":"660b5b0becc38ac2450f233d","name":"Rio Yokota","hidden":false},{"_id":"660b5b0becc38ac2450f233e","name":"Niklas Muennighoff","hidden":false},{"_id":"660b5b0becc38ac2450f233f","name":"Suhas Pai","hidden":false},{"_id":"660b5b0becc38ac2450f2340","name":"Tosin Adewumi","hidden":false},{"_id":"660b5b0becc38ac2450f2341","name":"Veronika Laippala","hidden":false},{"_id":"660b5b0becc38ac2450f2342","user":{"_id":"61919e59d221f4281b3833d5","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1669164541734-61919e59d221f4281b3833d5.jpeg","isPro":false,"fullname":"Xiaozhe Yao","user":"xzyao","type":"user"},"name":"Xiaozhe Yao","status":"claimed_verified","statusLastChangedAt":"2024-04-03T09:25:59.978Z","hidden":false},{"_id":"660b5b0becc38ac2450f2343","name":"Adalberto Junior","hidden":false},{"_id":"660b5b0becc38ac2450f2344","user":{"_id":"63972847b3e2256c9ce1307b","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63972847b3e2256c9ce1307b/Zt-O-F_gl0dDefbkOORfY.png","isPro":false,"fullname":"Alpay Ariyak","user":"alpayariyak","type":"user"},"name":"Alpay Ariyak","status":"claimed_verified","statusLastChangedAt":"2024-04-03T09:25:39.066Z","hidden":false},{"_id":"660b5b0becc38ac2450f2345","name":"Aleksandr Drozd","hidden":false},{"_id":"660b5b0becc38ac2450f2346","user":{"_id":"608995d619137b3a6ba76113","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/608995d619137b3a6ba76113/QK25kNi5XIlw8m3104KC8.jpeg","isPro":false,"fullname":"Jordan Clive","user":"jordiclive","type":"user"},"name":"Jordan Clive","status":"claimed_verified","statusLastChangedAt":"2024-04-02T11:05:14.866Z","hidden":false},{"_id":"660b5b0becc38ac2450f2347","user":{"_id":"625daa0c943e346492b63d21","avatarUrl":"/avatars/860f8e9bff85325096b6f8e15102def6.svg","isPro":false,"fullname":"Kshitij Gupta","user":"kshitijkg","type":"user"},"name":"Kshitij Gupta","status":"claimed_verified","statusLastChangedAt":"2024-04-02T11:03:01.538Z","hidden":false},{"_id":"660b5b0becc38ac2450f2348","name":"Liangyu Chen","hidden":false},{"_id":"660b5b0becc38ac2450f2349","user":{"_id":"644b983f0fbe4830f192c4f5","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/644b983f0fbe4830f192c4f5/4eX4OS0gnXGqLu3eaMml_.png","isPro":false,"fullname":"Qi Sun","user":"lfsm","type":"user"},"name":"Qi Sun","status":"claimed_verified","statusLastChangedAt":"2024-08-05T09:02:30.547Z","hidden":false},{"_id":"660b5b0becc38ac2450f234a","user":{"_id":"60e50ce5350d181892d5a636","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/60e50ce5350d181892d5a636/qjYikZtE-2kjub-6XG5Kn.jpeg","isPro":false,"fullname":"Ken Tsui","user":"kenhktsui","type":"user"},"name":"Ken Tsui","status":"claimed_verified","statusLastChangedAt":"2024-04-02T19:35:06.873Z","hidden":false},{"_id":"660b5b0becc38ac2450f234b","name":"Noah Persaud","hidden":false},{"_id":"660b5b0becc38ac2450f234c","name":"Nour Fahmy","hidden":false},{"_id":"660b5b0becc38ac2450f234d","name":"Tianlong Chen","hidden":false},{"_id":"660b5b0becc38ac2450f234e","name":"Mohit Bansal","hidden":false},{"_id":"660b5b0becc38ac2450f234f","user":{"_id":"62f3f80d4794a3c919bca0d5","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/62f3f80d4794a3c919bca0d5/R-OuVhyE5KwvuHx9hUhfu.png","isPro":true,"fullname":"Nicolò Monti","user":"nicolomonti","type":"user"},"name":"Nicolo Monti","status":"claimed_verified","statusLastChangedAt":"2024-04-02T19:35:13.690Z","hidden":false},{"_id":"660b5b0becc38ac2450f2350","name":"Tai Dang","hidden":false},{"_id":"660b5b0becc38ac2450f2351","user":{"_id":"6090ff099a8bcaa437b234a4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6090ff099a8bcaa437b234a4/iUvw7JXT-ngI7rGk1x-io.jpeg","isPro":false,"fullname":"Ziyang Luo","user":"Ziyang","type":"user"},"name":"Ziyang Luo","status":"claimed_verified","statusLastChangedAt":"2024-04-02T19:35:09.791Z","hidden":false},{"_id":"660b5b0becc38ac2450f2352","name":"Tien-Tung Bui","hidden":false},{"_id":"660b5b0becc38ac2450f2353","name":"Roberto Navigli","hidden":false},{"_id":"660b5b0becc38ac2450f2354","user":{"_id":"604eb19e3050a33ebb17ef58","avatarUrl":"/avatars/23650002ba3befee83060fe978a251c8.svg","isPro":false,"fullname":"Virendra Mehta","user":"Veeru","type":"user"},"name":"Virendra Mehta","status":"claimed_verified","statusLastChangedAt":"2024-09-23T13:07:20.289Z","hidden":false},{"_id":"660b5b0becc38ac2450f2355","name":"Matthew Blumberg","hidden":false},{"_id":"660b5b0becc38ac2450f2356","user":{"_id":"627bda4f1b02d7b5bf63f1f8","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/627bda4f1b02d7b5bf63f1f8/sqK5uRemxbv-zBQ9B8ugE.jpeg","isPro":false,"fullname":"VM","user":"cabbage972","type":"user"},"name":"Victor May","status":"claimed_verified","statusLastChangedAt":"2024-04-03T09:25:24.578Z","hidden":false},{"_id":"660b5b0becc38ac2450f2357","user":{"_id":"5fc6879e1c5ee87b1164876d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/5fc6879e1c5ee87b1164876d/Tjnm_lv0Bq0gPbFOTDH6E.jpeg","isPro":false,"fullname":"Huu Nguyen","user":"huu-ontocord","type":"user"},"name":"Huu Nguyen","status":"extracted_confirmed","statusLastChangedAt":"2024-04-02T02:11:32.098Z","hidden":false},{"_id":"660b5b0becc38ac2450f2358","user":{"_id":"602fb2d93d343696d96ef287","avatarUrl":"/avatars/a78ce59506a416e3df1d283a13e5aefa.svg","isPro":false,"fullname":"Sampo Pyysalo","user":"spyysalo","type":"user"},"name":"Sampo Pyysalo","status":"extracted_confirmed","statusLastChangedAt":"2024-04-02T08:08:30.308Z","hidden":false}],"publishedAt":"2024-03-30T15:38:54.000Z","submittedOnDailyAt":"2024-04-02T14:08:02.668Z","title":"Aurora-M: The First Open Source Multilingual Language Model Red-teamed\n according to the U.S. Executive Order","submittedOnDailyBy":{"_id":"60f1abe7544c2adfd699860c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674929746905-60f1abe7544c2adfd699860c.jpeg","isPro":false,"fullname":"AK","user":"akhaliq","type":"user"},"summary":"Pretrained language models underpin several AI applications, but their high\ncomputational cost for training limits accessibility. Initiatives such as BLOOM\nand StarCoder aim to democratize access to pretrained models for collaborative\ncommunity development. However, such existing models face challenges: limited\nmultilingual capabilities, continual pretraining causing catastrophic\nforgetting, whereas pretraining from scratch is computationally expensive, and\ncompliance with AI safety and development laws. This paper presents Aurora-M, a\n15B parameter multilingual open-source model trained on English, Finnish,\nHindi, Japanese, Vietnamese, and code. Continually pretrained from\nStarCoderPlus on 435 billion additional tokens, Aurora-M surpasses 2 trillion\ntokens in total training token count. It is the first open-source multilingual\nmodel fine-tuned on human-reviewed safety instructions, thus aligning its\ndevelopment not only with conventional red-teaming considerations, but also\nwith the specific concerns articulated in the Biden-Harris Executive Order on\nthe Safe, Secure, and Trustworthy Development and Use of Artificial\nIntelligence. Aurora-M is rigorously evaluated across various tasks and\nlanguages, demonstrating robustness against catastrophic forgetting and\noutperforming alternatives in multilingual settings, particularly in safety\nevaluations. To promote responsible open-source LLM development, Aurora-M and\nits variants are released at\nhttps://huggingface.co/collections/aurora-m/aurora-m-models-65fdfdff62471e09812f5407 .","upvotes":42,"discussionId":"660b5b0cecc38ac2450f237f","ai_summary":"Aurora-M, a 15B parameter multilingual open-source model trained on multiple languages and code, addresses challenges of catastrophic forgetting and AI safety through continual pretraining and human-reviewed safety instructions.","ai_keywords":["pretrained language models","multilingual capabilities","continual pretraining","catastrophic forgetting","open-source model","StarCoderPlus","human-reviewed safety instructions","Biden-Harris Executive Order","robustness","safety evaluations"]},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"643006f01572f43a481766a9","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/643006f01572f43a481766a9/egLlzmsWxHovmuvqAPnBO.jpeg","isPro":false,"fullname":"_","user":"Xa9aX","type":"user"},{"_id":"6308c49c454dc257521bc7f9","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6308c49c454dc257521bc7f9/UWUS6OPa6OpVu1T0gd-wJ.jpeg","isPro":false,"fullname":"Taishi","user":"Taishi-N324","type":"user"},{"_id":"60e50ce5350d181892d5a636","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/60e50ce5350d181892d5a636/qjYikZtE-2kjub-6XG5Kn.jpeg","isPro":false,"fullname":"Ken Tsui","user":"kenhktsui","type":"user"},{"_id":"60535c9d10aba34e3b6a2ef7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1616075850230-noauth.jpeg","isPro":false,"fullname":"vumichien","user":"vumichien","type":"user"},{"_id":"63972847b3e2256c9ce1307b","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63972847b3e2256c9ce1307b/Zt-O-F_gl0dDefbkOORfY.png","isPro":false,"fullname":"Alpay Ariyak","user":"alpayariyak","type":"user"},{"_id":"602fb2d93d343696d96ef287","avatarUrl":"/avatars/a78ce59506a416e3df1d283a13e5aefa.svg","isPro":false,"fullname":"Sampo Pyysalo","user":"spyysalo","type":"user"},{"_id":"60e3010bdc3e86876e16eaea","avatarUrl":"/avatars/6802da2521be787803f6f2f439a5616a.svg","isPro":true,"fullname":"Tosin Adewumi","user":"tosin","type":"user"},{"_id":"61b85aa99ba538c73a7dc78b","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/61b85aa99ba538c73a7dc78b/gWxtQAvOYn7cXgE_nAy0p.jpeg","isPro":false,"fullname":"Simone Tedeschi","user":"sted97","type":"user"},{"_id":"5dd96eb166059660ed1ee413","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/5dd96eb166059660ed1ee413/NQtzmrDdbG0H8qkZvRyGk.jpeg","isPro":true,"fullname":"Julien Chaumond","user":"julien-c","type":"user"},{"_id":"6340651b388c3fa40f9a5bc0","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6340651b388c3fa40f9a5bc0/av1C4_S7bHGxAzOu8lOmG.jpeg","isPro":false,"fullname":"Adam Molnar","user":"adamm-hf","type":"user"},{"_id":"6093a02dc4a92d63a91c5236","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6093a02dc4a92d63a91c5236/yUte6V0FU0BvVFAbON-9n.jpeg","isPro":true,"fullname":"Diwank Tomer","user":"diwank","type":"user"},{"_id":"62f3f80d4794a3c919bca0d5","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/62f3f80d4794a3c919bca0d5/R-OuVhyE5KwvuHx9hUhfu.png","isPro":true,"fullname":"Nicolò Monti","user":"nicolomonti","type":"user"}],"acceptLanguages":["*"],"dailyPaperRank":1}">
Aurora-M, a 15B parameter multilingual open-source model trained on multiple languages and code, addresses challenges of catastrophic forgetting and AI safety through continual pretraining and human-reviewed safety instructions.
AI-generated summary
Pretrained language models underpin several AI applications, but their high
computational cost for training limits accessibility. Initiatives such as BLOOM
and StarCoder aim to democratize access to pretrained models for collaborative
community development. However, such existing models face challenges: limited
multilingual capabilities, continual pretraining causing catastrophic
forgetting, whereas pretraining from scratch is computationally expensive, and
compliance with AI safety and development laws. This paper presents Aurora-M, a
15B parameter multilingual open-source model trained on English, Finnish,
Hindi, Japanese, Vietnamese, and code. Continually pretrained from
StarCoderPlus on 435 billion additional tokens, Aurora-M surpasses 2 trillion
tokens in total training token count. It is the first open-source multilingual
model fine-tuned on human-reviewed safety instructions, thus aligning its
development not only with conventional red-teaming considerations, but also
with the specific concerns articulated in the Biden-Harris Executive Order on
the Safe, Secure, and Trustworthy Development and Use of Artificial
Intelligence. Aurora-M is rigorously evaluated across various tasks and
languages, demonstrating robustness against catastrophic forgetting and
outperforming alternatives in multilingual settings, particularly in safety
evaluations. To promote responsible open-source LLM development, Aurora-M and
its variants are released at
https://huggingface.co/collections/aurora-m/aurora-m-models-65fdfdff62471e09812f5407 .