← 首页 | 厂商论文 | 导读
Meta

LLaMA: Open and Efficient Foundation Language Models

LLaMA:开放高效的基础语言模型

📅 2023-02-24👤 Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martin📄 arXiv: 2302.13971
LLaMA开源基础模型参数效率Meta

中文摘要

LLaMA(Large Language Model Meta AI)是Meta开发的开源基础语言模型系列,涵盖7B到65B参数规模。LLaMA在较小数据集上训练,但通过高效的训练策略实现了与更大模型相当的性能。7B参数的LLaMA在多项基准测试上超越GPT-3(175B参数),展现了参数效率的重要突破。

LLaMA (Large Language Model Meta AI) is Meta's open-source foundation language model series ranging from 7B to 65B parameters. LLaMA is trained on a smaller dataset but achieves competitive performance with larger models through efficient training strategies. The 7B parameter model outperforms GPT-3 (175B) on several benchmarks.

快速链接

← 厂商论文列表首页