0 citations
LLaMA: Open and Efficient Foundation Language Models
arXiv (Cornell University)2023
Citations Over Time
Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Édouard Grave, Guillaume Lample
Abstract
We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community.
Related Papers
- → Evaluation of time-based ranging methods: Does the choice matter?(2017)13 cited
- → Performance Comparison of Asynchronous Ranging Algorithms(2009)13 cited
- → Ranging-Aided Ground Robot Navigation Using UWB Nodes at Unknown Locations(2022)4 cited
- A laser ranging method with high precision and large range in high speed based on phase measurement(2015)
- → Performance Analysis of Ranging Techniques for the KPLO Mission(2018)