第一项始终是图钉的来源。总体可信度是各份资料对上方所用开始和结束时间支持程度的加权平均;资料每比最新的一份旧 180 天,权重减半。
arXiv 上 1706.03762 的元数据给出 citation_date 为 2017/06/12,维基百科关于该论文的条目也如此记载:“2017-06-12,最初的(1 亿参数)编码器—解码器 Transformer 模型在《Attention is all you need》论文中发表。”
Dates the paper to 12 June 2017 and describes what it did: "a 2017 research paper on machine learning authored by eight scientists and engineers working at Google" introducing the transformer, "based on the attention mechanism proposed in 2014 by Bahdanau et al.", which "has become the main architecture of a wide variety of artificial intelligence systems, including large language models". It also records that all eight authors later left Google.[1]
The architecture itself, for how self-attention replaced recurrence and what was built on it afterwards.[1]
有遗漏或错误吗?用你自己的话说明:能佐证此图钉的链接、不同的开始或结束日期及理由,或缺失、有误的信息。AI 会对照此图钉的来源进行核实,搜索更好的来源,并添加任何支持你说法的页面。图钉自身的来源仍然最重要。AI 也会查看图片:显示的是别的东西或显示效果差的图片会被移到后面或替换。
原标题: Google Publishes Attention Is All You Need