Daily Arxiv

Cette page résume et organise les publications en intelligence artificielle du monde entier.
Les contenus sont synthétisés grâce à Google Gemini et le service est proposé à but non lucratif.
Les droits d'auteur des articles appartiennent à leurs auteurs ou institutions respectives ; en cas de partage, il suffit d'en mentionner la source.

Modèles linguistiques de l'Apple Intelligence Foundation : rapport technique 2025

Created by
  • Haebom

Auteur

Ethan Li (Taoyi), Anders Boesen Lindbo Larsen (Taoyi), Chen Zhang (Taoyi), Xiyou Zhou (Taoyi), Jun Qin (Taoyi), Dian Ang Yap (Taoyi), Narendran Raghavan (Taoyi), Xuankai Chang (Taoyi), Margit Bowler (Taoyi), Eray Yildiz (Taoyi), John Peebles (Taoyi), Hannah Gillis Coleman (Taoyi), Matteo Ronchi (Taoyi), Peter Gray (Taoyi), Keen You (Taoyi), Anthony Spalvieri-Kruse (Taoyi), Ruoming Pang (Taoyi), Reed Li (Taoyi), Yuli Yang (Taoyi), Emad Soroush (Taoyi), Zhiyun Lu (Taoyi), Crystal Xiao (Taoyi), Rong Situ (Taoyi), Jordan Huffaker (Taoyi), David Griffiths (Taoyi), Zaïd Ahmed (Taoyi), Peng Zhang (Taoyi), Daniel Parilla (Taoyi), Asaf Liberman (Taoyi), Jennifer Mallalieu (Taoyi), Parsa Mazaheri (Taoyi), Qibin Chen (Taoyi), Manjot Bilkhu (Taoyi), Aonan Zhang (Taoyi), Eric Wang (Taoyi), Dave Nelson (Taoyi), Michael FitzMaurice (Taoyi), Thomas Voice (Taoyi), Jeremy Liu (Taoyi), Josh Shaffer (Taoyi), Shiwen Zhao (Taoyi), Prasanth Yadla (Taoyi), Farzin Rasteh (Taoyi), Pengsheng Guo (Taoyi), Arsalan Farooq (Taoyi), Jeremy Snow (Taoyi), Stephen Murphy (Taoyi), Tao Lei (Taoyi), Minsik Cho (Taoyi), George Horrell (Taoyi), Sam Dodge (Taoyi), Lindsay Hislop (Taoyi), Sumeet Singh (Taoyi), Alex Dombrowski (Taoyi), Aiswarya Raghavan (Taoyi), Sasha Sirovica (Taoyi), Mandana Saebi (Taoyi), Faye Lao (Taoyi), Max Lam (Taoyi), TJ Lu (Taoyi), Zhaoyang Xu (Taoyi), Karanjeet Singh (Taoyi), Marc Kirchner (Taoyi), David Mizrahi (Taoyi), Rajat Arora (Taoyi), Haotian Zhang (Taoyi), Henry Mason (Taoyi), Lawrence Zhou (Taoyi), Yi Hua (Taoyi), Ankur Jain (Taoyi), Felix Bai (Taoyi), Joseph Astrauskas (Taoyi), Floris Weers (Taoyi), Josh Gardner (Taoyi), Mira Chiang (Taoyi), Yi Zhang (Taoyi), Pulkit Agrawal (Taoyi), Tony Sun (Taoyi), Quentin Keunebroek (Taoyi), Matthew Hopkins (Taoyi), Bugu Wu (Taoyi), Tao Jia (Taoyi), Chen Chen (Taoyi), Zirui Wang (Taoyi), Cha Chen (Taoyi), Xiaoyi Ren (Taoyi), Feng Nan (Taoyi), Josh Elman (Taoyi), Dong Yin (Taoyi), Yusuf Goren (Taoyi), Jeff Lai (Taoyi), Yiran Fei (Taoyi), Syd Evans (Taoyi), Muyang Yu (Taoyi), Guoli Yin (Taoyi), Yi Qin (Taoyi), Erin Feldman (Taoyi), Isha Garg (Taoyi), Aparna Rajamani (Taoyi), Karla Vega (Taoyi), Walker Cheng (Taoyi), TJ Collins (Taoyi), Hans Han (Taoyi), Raul Rea Menacho (Taoyi), Simon Yeung (Taoyi), Sophy Lee (Taoyi), Phani Mutyala (Taoyi), Ying-Chang Cheng (Taoyi), Zhe Gan (Taoyi), Sprite Chu (Taoyi), Justin Lazarow (Taoyi), Alessandro Pappalardo (Taoyi), Federico Scozzafava (Taoyi), Jing Lu (Taoyi), Erik Daxberger (Taoyi), Laurent Duchesne (Taoyi), Jen Liu (Taoyi), David Guera (Taoyi), Stefano Ligas (Taoyi), Mary Beth Kery (Taoyi), Brent Ramerth (Taoyi), Ciro Sannino (Taoyi), Marcin Eichner (Taoyi), Haoshuo Huang (Taoyi), Rui Qian (Taoyi), Moritz Schwarzer-Becker (Taoyi), David Riazati (Taoyi), Mingfei Gao (Taoyi), Bailin Wang (Taoyi), Jack Cackler (Taoyi), Yang Lu (Taoyi), Ransen Niu (Taoyi), John Dennison (Taoyi), Guillaume Klein (Taoyi), Jeffrey Bigham (Taoyi), Deepak Gopinath (Taoyi), Navid Shiee (Taoyi), Darren Botten (Taoyi), Guillaume Tartavel (Taoyi), Alex Guillen Garcia (Taoyi), Sam Xu (Taoyi), Victoria MonchJuan Haladjian (Taoyi), Zi-Yi Dou (Taoyi), Matthias Paulik (Taoyi), Adolfo Lopez Mendez (Taoyi), Zhen Li (Taoyi), Hong-You Chen (Taoyi), Chao Jia (Taoyi), Dhaval Doshi (Taoyi), Zhengdong Zhang (Taoyi), Raunak Manjani (Taoyi), Aaron Franklin (Taoyi), Zhile Ren (Taoyi), David Chen (Taoyi), Artsiom Peshko (Taoyi), Nandhitha Raghuram (Taoyi), Hans Hao (Taoyi), Jiulong Shan (Taoyi), Kavya Nerella (Taoyi), Ramsey Tantawi (Taoyi), Vivek Kumar (Taoyi), Saiwen Wang (Taoyi), Brycen Wershing (Taoyi), Bhuwan Dhingra (Taoyi), Dhruti Shah (Taoyi), Ob Adaranijo (Taoyi), Xin Zheng (Taoyi), Tait Madsen (Taoyi), Hadas Kotek (Taoyi), Chang Liu (Taoyi), Yin Xia (Taoyi), Hanli Li (Taoyi), Suma Jayaram (Taoyi), Yanchao Sun (Taoyi), Ahmed Fakhry (Taoyi), Vasileios Saveris (Taoyi), Dustin Withers (Taoyi), Yanghao Li (Taoyi), Alp Aygar (Taoyi), Andres Romero Mier Y Teran (Taoyi), Kaiwei Huang (Taoyi), Mark Lee (Taoyi), Xiujun Li (Taoyi), Yuhong Li (Taoyi), Tyler Johnson (Taoyi), Jay Tang (Taoyi), Joseph Yitan Cheng (Taoyi), Futang Peng (Taoyi), Andrew Walkingshaw (Taoyi), Lucas Guibert (Taoyi), Abhishek Sharma (Taoyi), Cheng Shen (Taoyi), Piotr Maj (Taoyi), Yasutaka Tanaka (Taoyi), You-Cyuan Jhang (Taoyi), Vivian Ma (Taoyi), Tommi Vehvilainen (Taoyi), Kelvin Zou (Taoyi), Jeff Nichols (Taoyi), Matthew Lei (Taoyi), David Qiu (Taoyi), Yihao Qian (Taoyi), Gokul Santhanam (Taoyi), Wentao Wu (Taoyi), Yena Han (Taoyi), Dominik Moritz (Taoyi), Haijing Fu (Taoyi), Mingze Xu (Taoyi), Vivek Rathod (Taoyi), Jian Liu (Taoyi), Louis D'hauwe (Taoyi), Qin Ba (Taoyi), Haitian Sun (Taoyi), Haoran Yan (Taoyi), Philipp Dufter (Taoyi), Anh Nguyen (Taoyi), Yihao Feng (Taoyi), Emma Wang (Taoyi), Keyu He (Taoyi), Rahul Nair (Taoyi), Sanskruti Shah (Taoyi), Jiarui Lu (Taoyi), Patrick Sonnenberg (Taoyi), Jeremy Warner (Taoyi), Yuanzhi Li (Taoyi), Bowen Pan (Taoyi), Ziyi Zhong (Taoyi), Joe Zhou (Taoyi), Sam Davarnia (Taoyi), Olli Saarikivi (Taoyi), Irina Belousova (Taoyi), Rachel Burger (Taoyi), Shang-Chen Wu (Taoyi), Di Feng (Taoyi), Bas Straathof (Taoyi), James Chou (Taoyi), Yuanyang Zhang (Taoyi), Marco Zuliani (Taoyi), Eduardo Jimenez (Taoyi), Abhishek Sundararajan (Taoyi), Xianzhi Du (Taoyi), Chang Lan (Taoyi), Nilesh Shahdadpuri (Taoyi), Peter Grasch (Taoyi), Sergiu Sima (Taoyi), Josh Newnham (Taoyi), Varsha Paidi (Taoyi), Jianyu Wang (Taoyi), Kaelen Haag (Taoyi), Alex Braunstein (Taoyi), Daniele Molinari (Taoyi), Richard Wei (Taoyi), Brenda Yang (Taoyi), Nicholas Lusskin (Taoyi), Joanna Arreaza-Taylor (Taoyi), Meng Cao (Taoyi), Nicholas Seidl (Taoyi), Simon Wang (Taoyi), Jiaming Hu (Taoyi), Yiping Ma (Taoyi), Mengyu Li (Taoyi), Kieran Liu (Taoyi), Hang Su (Taoyi), Sachin Ravi (Taoyi), Chong Wang (Taoyi), (Taoyi), Jane Li (Taoyi), Andrew Hansen (Taoyi), Javier Movellan (Taoyi), Christopher Neubauer (Taoyi), Hanzhi Zhou (Taoyi), Chris Chaney (Taoyi), Nazir Kamaldin (Taoyi), Valentin Wolf (Taoyi), Fernando Bermudez -Medina (Taoyi), Joris Pelemans (Taoyi), Peter Fu (Taoyi), Howard Xing (Taoyi), Xiang Kong (Taoyi), Wayne Shan (Taoyi), Gabriel Jacoby-Cooper (Taoyi), Dongcai Shen (Taoyi), Tom Gunter (Taoyi), Guillaume Seguin (Taoyi), Fangping Shi (Taoyi), Shiyu Li (Taoyi), Yang Xu (Taoyi), Areeba Kamal (Taoyi), Dan Masi (Taoyi), Saptarshi Guha (Taoyi), Qi Zhu (Taoyi), Jenna Thibodeau (Taoyi), Changyuan Zhang (Taoyi), Rebecca Callahan (Taoyi), Charles Maalouf (Taoyi), Wilson Tsao (Taoyi), Boyue Li (Taoyi), Qingqing Cao (Taoyi), Naomy Sabo (Taoyi), Cheng Leong (Taoyi), Yi Wang (Taoyi), Anupama Mann Anupama (Taoyi), Colorado Reed (Taoyi), Kenneth Jung (Taoyi), Zhifeng Chen (Taoyi), Mohana Prasad Sathya Moorthy (Taoyi), Yifei He (Taoyi), Erik Hornberger (Taoyi), Devi Krishna (Taoyi), Senyu Tong (Taoyi), Michael (Taoyi), Lee, David Haldimann, Yang Zhao, Bowen Zhang, Chang Gao, Chris Bartels, Sushma Rao, Nathalie Tran, Simon Lehnerer, Co Giang, Patrick Dong, Junting Pan, Biyao Wang, Dongxu Li, Mehrdad Farajtabar, Dongseong Hwang, Grace Duanmu, Eshan Verma, Sujeeth Reddy, Qi Shan, Hongbin Gao, Nan Du, Pragnya Sridhar, Forrest Huang, Yingbo Wang, Nikhil Bhendawade, Diane Zhu, Sai Aitharaju, Fred Hohman, Lauren Gardiner, Chung-Cheng Chiu, Yinfei Yang, Alper Kokmen, Frank Chu, Ke Ye, Kaan Elgin, Oron Levy, John Park, Donald Zhang, Eldon Schoop, Nina Wenzel, Michael Booker, Hyunjik Kim, Chinguun Erdenebileg, Nan Dun, Eric Liang Yang, Priyal Chhatrapati, Vishaal Mahtani, Haiming Gang, Kohen Chia, Deepa Seshadri, Donghan Yu, Yan Meng, Kelsey Peterson, Zhen Yang, Yongqiang Wang, Carina Peng, Doug Kang, Anuva Agarwal, Albert Antony, Juan Lao Tebar, Albin Madappally Jose, Regan Poston, Andy De Wang, Gerard Casamayor, Elmira Amirloo, Violet Yao, Wojciech Kryscinski, Kun Duan, Lezhi L

Contour

Apple présente deux modèles de langage multilingues et multimodaux qui optimisent les fonctionnalités d'Apple Intelligence sur les appareils et services Apple. Le premier est un modèle embarqué de 3 milliards de paramètres, optimisé pour Apple Silicon grâce à des innovations architecturales (partage de cache KV et apprentissage sensible à la quantification 2 bits). Le second est un modèle serveur évolutif basé sur un nouveau transformateur PT-MoE (Parallel Track Mixture of Experts) combinant parallélisme de pistes, calcul épars mixte expert-expert et attention globale-locale entrelacée. Les deux modèles sont entraînés sur des jeux de données multilingues et multimodaux à grande échelle issus d'une exploration web responsable, de corpus sous licence et de données synthétiques de haute qualité. Ils sont ensuite optimisés par un réglage fin supervisé et un apprentissage par renforcement sur une nouvelle plateforme asynchrone. Les modèles ainsi obtenus comprennent les images, exécutent des appels d'outils et prennent en charge plusieurs langues supplémentaires. Les performances des modèles serveur et embarqué sont comparables, voire supérieures, à celles de références ouvertes de taille similaire, lors de benchmarks publics et d'évaluations humaines. Un nouveau framework de modèle fondamental, centré sur Swift, permet la génération de guides, les appels d'outils contraints et le réglage des adaptateurs LoRA, permettant aux développeurs d'intégrer ces fonctionnalités en quelques lignes de code seulement. Les dernières avancées des modèles d'intelligence d'Apple s'appuient sur l'engagement d'Apple en faveur d'une IA responsable, avec des innovations comme le Cloud Compute privé et des mesures de protection comme le filtrage de contenu et les classifications régionales, le tout dans le respect de la confidentialité des utilisateurs.

Takeaways, Limitations

•
Takeaways:
◦
Développement d'un modèle de langage multilingue et multimodal hautes performances pour une application aux appareils et services Apple.
◦
Des innovations architecturales pour l'optimisation des modèles sur appareil (partage de cache KV, apprentissage sensible à la quantification 2 bits) sont présentées.
◦
Architecture de transformateur PT-MoE proposée pour les modèles de serveurs évolutifs.
◦
Approches d’IA responsables et protections de la vie privée renforcées.
◦
Accessibilité améliorée pour les développeurs grâce aux frameworks basés sur Swift.
◦
A démontré une performance compétitive dans les benchmarks publics et les évaluations humaines.
•
Limitations:
◦
Manque d’informations détaillées sur la taille ou la composition de l’ensemble de données spécifique dans l’article.
◦
Manque d'explication technique spécifique du Private Cloud Compute.
◦
Il manque des précisions sur une « approche responsable de l'IA ». Des explications détaillées sur les garanties et les méthodes d'évaluation spécifiques sont nécessaires.
◦
Manque de prise en compte de l'impact environnemental du modèle (consommation énergétique, etc.).
◦
Absence de mention des plans de maintenance et de mise à jour à long terme.
👍
Made with Slashpage