PaLM 2 Technical Report

作者：Rohan Anil, Andrew M. Dai, Orhan Firat, Melvin Johnson, Dmitry Lepikhin, Alexandre Passos, Siamak Shakeri, Emanuel Taropa, Paige Bailey, Z. Chen, Eric Chu, J. Clark, Laurent El Shafey, Yanping Huang, K. Meier-Hellstern, Gaurav Mishra, Erica Moreira, Mark Omernick, Kevin Robinson, Sebastian Ruder, Yi Tay, Kefan Xiao, Yuan-Zhong Xu, Yujing Zhang, Gustavo Hernández Abrego, Junwhan Ahn, Jacob Austin, P. Barham, Jan A. Botha, James Bradbury, Siddhartha Brahma, K. Brooks, Michele Catasta, Yong-Zhou Cheng, Colin Cherry, Christopher A. Choquette-Choo, A. Chowdhery, C. Crépy, Shachi Dave, Mostafa Dehghani, Sunipa Dev, Jacob Devlin, M. D'iaz, Nan Du, Ethan Dyer, Vladimir Feinberg, Fan Feng, Vlad Fienber, Markus Freitag, Xavier García, Sebastian Gehrmann, Lucas González, Guy Gur-Ari, S. Hand, Hadi Hashemi, Le Hou, Joshua Howland, A. Hu, Jeffrey Hui, Jeremy Hurwitz, M. Isard, Abe Ittycheriah, Matthew Jagielski, Wen-Hao Jia, Kathleen Kenealy, M. Krikun, Sneha Kudugunta, Chang Lan, Katherine Lee, Benjamin Lee, Eric Li, Mu-Li Li, Wei Li, Yaguang Li, Jun Yu Li, Hyeontaek Lim, Han Lin, Zhong-Zhong Liu, Frederick Liu, M. Maggioni, Aroma Mahendru, Joshua Maynez, Vedant Misra, Maysam Moussalem, Zachary Nado, John Nham, Eric Ni, A. Nystrom, Alicia Parrish, Marie Pellat, M. Polacek, Oleksandr Polozov, Reiner Pope, Siyuan Qiao, Emily Reif, Bryan Richter, Parker Riley, A. Ros, Aurko Roy, Brennan Saeta, R. Samuel, R. Shelby, Ambrose Slone, D. Smilkov, David R. So, Daniela Sohn, Simon Tokumine, Dasha Valter, Vijay Vasudevan, Kiran Vodrahalli, Xuezhi Wang, Pidong Wang, Zirui Wang, Tao Wang, J. Wieting, Yuhuai Wu, Ke Xu, Yunhan Xu, Linting Xue, Pengcheng Yin, Jiahui Yu, Qiaoling Zhang, Steven Zheng, Ce Zheng, Wei Zhou, Denny Zhou, Slav Petrov, Yonghui Wu · 发表于：arXiv.org · 年份：2023 · 被引用次数：1555 · 研究领域：Computer Science

We introduce PaLM 2, a new state-of-the-art language model that has better multilingual and reasoning capabilities and is more compute-efficient than its predecessor PaLM. PaLM 2 is a Transformer-based model trained using a mixture of objectives. Through extensive evaluations on English and multilingual language, and reasoning tasks, we demonstrate that PaLM 2 has significantly improved quality on downstream tasks across different model sizes, while simultaneously exhibiting faster and more efficient inference compared to PaLM. This improved efficiency enables broader deployment while also allowing the model to respond faster, for a more natural pace of interaction. PaLM 2 demonstrates robust reasoning capabilities exemplified by large improvements over PaLM on BIG-Bench and other reasoning tasks. PaLM 2 exhibits stable performance on a suite of responsible AI evaluations, and enables inference-time control over toxicity without additional overhead or impact on other capabilities. Overall, PaLM 2 achieves state-of-the-art performance across a diverse set of tasks and capabilities. When discussing the PaLM 2 family, it is important to distinguish between pre-trained models (of various sizes), fine-tuned variants of these models, and the user-facing products that use these models. In particular, user-facing products typically include additional pre- and post-processing steps. Additionally, the underlying models may evolve over time. Therefore, one should not expect the performa...