Talk:Attention Is All You Need

Shouldn't this article list the eight authors of the paper? If this is truly a landmark article in the field of artificial intelligence, seems very strange to say there are 8 authors but then never mention them.Tmb720 (talk) 17:29, 2 May 2024 (UTC) Reply

Seems to have been added before we started working on this article & Added some additional information about their specific contributions to the article as listed by the paper itself Akshithio (talk) 03:34, 2 December 2024 (UTC) Reply

I left the following feedback for the creator/future reviewers while reviewing this article: Good start

North8000 (talk) 21:20, 20 March 2024 (UTC) Reply

This article was the subject of a Wiki Education Foundation-supported course assignment, between 19 August 2024 and 7 December 2024. Further details are available on the course page . Student editor(s): Wangjx0 , Akshithio , Actuallyusedaccount , Rfheise , NanoRein (article contribs ). Peer reviewers: Aubreycape , Sgoellner03 , Skakkar4 .

— Assignment last updated by Skakkar4 (talk) 19:51, 29 November 2024 (UTC) Reply

The section transcluded from Transformer (deep learning architecture)#History contains some invocations of named references that are now showing up here as undefined. This seems to be a deficiency of the transclusion mechanism. 2.200.163.36 (talk) 03:05, 23 October 2024 (UTC) Reply

I will get someone at the Wikipedia:Help desk to look. Thanks for pointing it out. Commander Keane (talk) 04:19, 23 October 2024 (UTC) Reply

Done Folly Mox (talk) 11:31, 23 October 2024 (UTC) Reply

This differs from the original form of the Attention mechanism introduced in 2024. Seems wrong because the original Attention mechanism can not be made in 2024 when the paper was written in 2017, and the paper is based on this mechanism. 2001:16B8:C031:E300:1C2F:2566:BFB5:73DC (talk) 10:06, 9 December 2024 (UTC) Reply

Wow, how the heck was that even written out? Good catch! alby13 (talk) 21:43, 28 February 2025 (UTC) Reply

On February 28. 2025 I corrected the amount of times that the paper was cited according to the source news article as 80,000. On March 27, 2025, the user at the IP address 130.18.104.146 - https://en.wikipedia.org/wiki/Special:Contributions/130.18.104.146 -changed the figure to the incorrect 173,000. I would like this to be the correct information, so I hope that this will lead to it being corrected and it not being edited again. — Preceding unsigned comment added by Alby13 (talk • contribs) 20:33, 30 March 2025 (UTC) Reply

Generally one thing that readers would look for in an article about a research paper is where the paper was published. Was this paper published in a formal journal, or has it been peer-reviewed? It doesn't look like it, but maybe I'm missing something obvious. The cited Nature article about its popularity is ambiguous, but it does emphasize how common pre-prints are in AI research. It seems to assume that such papers are eventually published. It doesn't appear that Google Research has clear editorial standards or provides any info on peer review. Grayfell (talk) 00:51, 1 December 2025 (UTC) Reply

The T variable in the Attention Equation is not defined. — Preceding unsigned comment added by ~2026-15294-64 (talk) 19:38, 10 March 2026 (UTC) Reply

T is not a variable, it's the transpose. I.e., Q x K^T computes the scalar product between each query-key pair in matrix form. ~2026-41295-85 (talk) 11:37, 23 July 2026 (UTC) Reply

"After the paper, each of the authors left Google to join other companies or to found startups." appears twice in the first couple of paragraphs. --Treekids (talk) 17:42, 15 April 2026 (UTC) Reply