{"paper":{"title":"Aurora-M: Open Source Continual Pre-training for Multilingual Language and Code","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Adalberto Junior, Aleksandr Drozd, Alpay Ariyak, Arnav Varma Dantuluri, Ben Bogin, Diganta Misra, Felix Friedrich, Huu Nguyen, Jason T Stillerman, Jordan Clive, Ken Tsui, Kshitij Gupta, Liangyu Chen, Marzena Karpinska, Matthew Blumberg, Mayank Mishra, Mohit Bansal, Nicolo Monti, Niklas Muennighoff, Noah Persaud, Nour Fahmy, Prateek Yadav, Qi Sun, Rio Yokota, Roberto Navigli, Sampo Pyysalo, Simone Tedeschi, Suhas Pai, Tai Dang, Taishi Nakamura, Tanmay Laud, Terry Yue Zhuo, Tianlong Chen, Tien-Tung Bui, Tommaso Furlanello, Tosin Adewumi, Veronika Laippala, Victor May, Virendra Mehta, Vu Minh Chien, Wojciech Kusa, Xiaozhe Yao, Xuan-Son Vu, Yekun Chai, Ziyang Luo","submitted_at":"2024-03-30T15:38:54Z","abstract_excerpt":"Pretrained language models are an integral part of 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. Despite these efforts, such models encounter challenges such as limited multilingual capabilities, risks of catastrophic forgetting during continual pretraining, and the high costs of training models from scratch, alongside the need to align with AI safety standards and regulatory frameworks.\n  This paper presents Aurora-M, a 15B par"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.00399","kind":"arxiv","version":3},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2404.00399/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}