학술논문

BLOOM: A 176B-Parameter Open-Access Multilingual Language Model
Document Type
Working Paper
Author
Workshop, BigScienceScao, Teven LeFan, AngelaAkiki, ChristopherPavlick, EllieIlić, SuzanaHesslow, DanielCastagné, RomanLuccioni, Alexandra SashaYvon, FrançoisGallé, MatthiasTow, JonathanRush, Alexander M.Biderman, StellaWebson, AlbertAmmanamanchi, Pawan SasankaWang, ThomasSagot, BenoîtMuennighoff, Niklasdel Moral, Albert VillanovaRuwase, OlatunjiBawden, RachelBekman, StasMcMillan-Major, AngelinaBeltagy, IzNguyen, HuuSaulnier, LucileTan, SamsonSuarez, Pedro OrtizSanh, VictorLaurençon, HugoJernite, YacineLaunay, JulienMitchell, MargaretRaffel, ColinGokaslan, AaronSimhi, AdiSoroa, AitorAji, Alham FikriAlfassy, AmitRogers, AnnaNitzav, Ariel KreisbergXu, CanwenMou, ChenghaoEmezue, ChrisKlamm, ChristopherLeong, Colinvan Strien, DanielAdelani, David IfeoluwaRadev, DragomirPonferrada, Eduardo GonzálezLevkovizh, EfratKim, EthanNatan, Eyal BarDe Toni, FrancescoDupont, GérardKruszewski, GermánPistilli, GiadaElsahar, HadyBenyamina, HamzaTran, HieuYu, IanAbdulmumin, IdrisJohnson, IsaacGonzalez-Dios, Itziarde la Rosa, JavierChim, JennyDodge, JesseZhu, JianChang, JonathanFrohberg, JörgTobing, JosephBhattacharjee, JoydeepAlmubarak, KhalidChen, KimboLo, KyleVon Werra, LeandroWeber, LeonPhan, Longallal, Loubna BenTanguy, LudovicDey, MananMuñoz, Manuel RomeroMasoud, MaraimGrandury, MaríaŠaško, MarioHuang, MaxCoavoux, MaximinSingh, MayankJiang, Mike Tian-JianVu, Minh ChienJauhar, Mohammad A.Ghaleb, MustafaSubramani, NishantKassner, NoraKhamis, NurulaqillaNguyen, OlivierEspejel, Omarde Gibert, OnaVillegas, PauloHenderson, PeterColombo, PierreAmuok, PriscillaLhoest, QuentinHarliman, RhezaBommasani, RishiLópez, Roberto LuisRibeiro, RuiOsei, SalomeyPyysalo, SampoNagel, SebastianBose, ShamikMuhammad, Shamsuddeen HassanSharma, ShanyaLongpre, ShayneNikpoor, SomaiehSilberberg, StanislavPai, SuhasZink, SydneyTorrent, Tiago TimponiSchick, TimoThrush, TristanDanchev, ValentinNikoulina, VassilinaLaippala, VeronikaLepercq, ViolettePrabhu, VrindaAlyafeai, ZaidTalat, ZeerakRaja, ArunHeinzerling, BenjaminSi, ChengleiTaşar, Davut EmreSalesky, ElizabethMielke, Sabrina J.Lee, Wilson Y.Sharma, AbheeshtSantilli, AndreaChaffin, AntoineStiegler, ArnaudDatta, DebajyotiSzczechla, ElizaChhablani, GunjanWang, HanPandey, HarshitStrobelt, HendrikFries, Jason AlanRozen, JosGao, LeoSutawika, LintangBari, M SaifulAl-shaibani, Maged S.Manica, MatteoNayak, NihalTeehan, RyanAlbanie, SamuelShen, ShengBen-David, SrulikBach, Stephen H.Kim, TaewoonBers, TaliFevry, ThibaultNeeraj, TrishalaThakker, UrmishRaunak, VikasTang, XiangruYong, Zheng-XinSun, ZhiqingBrody, ShakedUri, YallowTojarieh, HadarRoberts, AdamChung, Hyung WonTae, JaesungPhang, JasonPress, OfirLi, ConglongNarayanan, DeepakBourfoune, HatimCasper, JaredRasley, JeffRyabinin, MaxMishra, MayankZhang, MinjiaShoeybi, MohammadPeyrounette, MyriamPatry, NicolasTazi, NouamaneSanseviero, Omarvon Platen, PatrickCornette, PierreLavallée, Pierre FrançoisLacroix, RémiRajbhandari, SamyamGandhi, SanchitSmith, ShadenRequena, StéphanePatil, SurajDettmers, TimBaruwa, AhmedSingh, AmanpreetCheveleva, AnastasiaLigozat, Anne-LaureSubramonian, ArjunNévéol, AurélieLovering, CharlesGarrette, DanTunuguntla, DeepakReiter, EhudTaktasheva, EkaterinaVoloshina, EkaterinaBogdanov, EliWinata, Genta IndraSchoelkopf, HaileyKalo, Jan-ChristophNovikova, JekaterinaForde, Jessica ZosaClive, JordanKasai, JungoKawamura, KenHazan, LiamCarpuat, MarineClinciu, MirunaKim, NajoungCheng, NewtonSerikov, OlegAntverg, Omervan der Wal, OskarZhang, RuiZhang, RuochenGehrmann, SebastianMirkin, ShacharPais, ShaniShavrina, TatianaScialom, ThomasYun, TianLimisiewicz, TomaszRieser, VerenaProtasov, VitalyMikhailov, VladislavPruksachatkun, YadaBelinkov, YonatanBamberger, ZacharyKasner, ZdeněkRueda, AlicePestana, AmandaFeizpour, AmirKhan, AmmarFaranak, AmySantos, AnaHevia, AnthonyUnldreaj, AntigonaAghagol, ArashAbdollahi, ArezooTammour, AychaHajiHosseini, AzadehBehroozi, BaharehAjibade, BenjaminSaxena, BharatFerrandis, Carlos MuñozMcDuff, DanielContractor, DanishLansky, DavidDavid, DavisKiela, DouweNguyen, Duong A.Tan, EdwardBaylor, EmiOzoani, EzinwanneMirza, FatimaOnoniwu, FranklineRezanejad, HabibJones, HessieBhattacharya, IndraniSolaiman, IreneSedenko, IrinaNejadgholi, IsarPassmore, JesseSeltzer, JoshSanz, Julio BonisDutra, LiviaSamagaio, MaironElbadri, MaraimMieskes, MargotGerchick, MarissaAkinlolu, MarthaMcKenna, MichaelQiu, MikeGhauri, MuhammedBurynok, MykolaAbrar, NafisRajani, NazneenElkott, NourFahmy, NourSamuel, OlanrewajuAn, RanKromann, RasmusHao, RyanAlizadeh, SamiraShubber, SarmadWang, SilasRoy, SouravViguier, SylvainLe, ThanhOyebade, TobiLe, TrieuYang, YoyoNguyen, ZachKashyap, Abhinav RameshPalasciano, AlfredoCallahan, AlisonShukla, AnimaMiranda-Escalada, AntonioSingh, AyushBeilharz, BenjaminWang, BoBrito, CaioZhou, ChenxiJain, ChiragXu, ChuxinFourrier, ClémentinePeriñán, Daniel LeónMolano, DanielYu, DianManjavacas, EnriqueBarth, FabioFuhrimann, FlorianAltay, GabrielBayrak, GiyaseddinBurns, GullyVrabec, Helena U.Bello, ImaneDash, IshaniKang, JihyunGiorgi, JohnGolde, JonasPosada, Jose DavidSivaraman, Karthik RangasaiBulchandani, LokeshLiu, LuShinzato, Luisade Bykhovetz, Madeleine HahnTakeuchi, MaikoPàmies, MarcCastillo, Maria ANezhurina, MariannaSänger, MarioSamwald, MatthiasCullan, MichaelWeinberg, MichaelDe Wolf, MichielMihaljcic, MinaLiu, MinnaFreidank, MoritzKang, MyungsunSeelam, NatashaDahlberg, NathanBroad, Nicholas MichioMuellner, NikolausFung, PascaleHaller, PatrickChandrasekhar, RamyaEisenberg, RenataMartin, RobertCanalli, RodrigoSu, RosalineSu, RuisiCahyawijaya, SamuelGarda, SamueleDeshmukh, Shlok SMishra, ShubhanshuKiblawi, SidOtt, SimonSang-aroonsiri, SineeKumar, SrishtiSchweter, StefanBharati, SushilLaud, TanmayGigant, ThéoKainuma, TomoyaKusa, WojciechLabrak, YanisBajaj, Yash ShaileshVenkatraman, YashXu, YifanXu, YingxinXu, YuTan, ZheXie, ZhongliYe, ZifanBras, MathildeBelkada, YounesWolf, Thomas
Source
Subject
Computer Science - Computation and Language
Language
Abstract
Large language models (LLMs) have been shown to be able to perform new tasks based on a few demonstrations or natural language instructions. While these capabilities have led to widespread adoption, most LLMs are developed by resource-rich organizations and are frequently kept from the public. As a step towards democratizing this powerful technology, we present BLOOM, a 176B-parameter open-access language model designed and built thanks to a collaboration of hundreds of researchers. BLOOM is a decoder-only Transformer language model that was trained on the ROOTS corpus, a dataset comprising hundreds of sources in 46 natural and 13 programming languages (59 in total). We find that BLOOM achieves competitive performance on a wide variety of benchmarks, with stronger results after undergoing multitask prompted finetuning. To facilitate future research and applications using LLMs, we publicly release our models and code under the Responsible AI License.