| This is an archive of past discussions on Wikipedia:WikiProject AI Cleanup/Noticeboard. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current main page. |
| Archive 1 | Archive 2 | Archive 3 | Archive 4 |
Possible LLM usage by User:Polymath Want To Be
The following discussion is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.
Polymath Want To Be (talk · contribs · deleted contribs · logs · filter log · block user · block log)
Asked them to LLMDISCLOSE on User talk:Polymath Want To Be § LLM disclosure but they denied it, even though they have inserted references with utm_source=chatgpt.com at least four times:
All those edits are to List of African-American U.S. state firsts (most of their edits seem to be to that one article and Timeline of African-American firsts). Because this is a timeline-style article I have some trouble with parsing it for WP:AISIGNS, and would appreciate if someone else could take a look. --Gurkubondinn (talk) 22:08, 3 December 2025 (UTC)
- Yeah I've seen them repeatedly trigger 1346 (hist · log) but havent done anything because they aren't adding much text. Have you checked to see if you can find any source-to-text integrity issues? NicheSports (talk) 22:21, 3 December 2025 (UTC)
- That's where I came across them, I've tagged the article and added
{{AI-generated source?}}inline with the questionable refs. - I have only found small source-to-text issues so far. For example, in Diff/1323483361 the ref sets
|publisher=Michigan Chronicalbut the correct name is "Michigan Chronicle" (and it should use the|work=parameter). This honestly seems more like a human mistake to me, but how else do you end up with that UTM parameter? - Others are harder for me to verify, Diff/1323485306 cites thegrio
.com which blocks me with a 403 from Cloudflare, presumably because I'm not in the US. The page is archived on archive.org, and while it confirms the claim that it is used as a source for, the author credited in the reference does not appear on the archived copy (dated April 2015) of the page that I am looking at. --Gurkubondinn (talk) 22:41, 3 December 2025 (UTC)/2012 /06 /07 /natasha-trethewey-african-american-woman-named-19th-u-s-poet-laureate / - {{AI-generated source?}} is intended for hallucinated sources. There is a more specific inline tag of {{AI-retrieved source}} for when sources exist but were found using an AI (utm_source refs), after such a a source is checked
|checked=can be added to the template to hide it. fifteen thousand two hundred twenty four (talk) 21:24, 4 December 2025 (UTC)- Ugh, thanks. Got those two mixed up, I'll correct the tags. Thanks for pointing it out to me. --Gurkubondinn (talk) 21:31, 4 December 2025 (UTC)
- {{AI-generated source?}} is intended for hallucinated sources. There is a more specific inline tag of {{AI-retrieved source}} for when sources exist but were found using an AI (utm_source refs), after such a a source is checked
- I spot checked these yesterday and didn't find anything drastic. The biggest one I found is this diff; the source exists and does mention her on the page stated, but not all of the facts added in the edit are actually mentioned in the source. That said, they are true (and are in our Helen Elizabeth Nash article already), which is already much better than most of what we review here. Gnomingstuff (talk) 15:17, 4 December 2025 (UTC)
- That's where I came across them, I've tagged the article and added
- This user is also doing a fair amount of reference gnoming on pages that might not otherwise get any attention, so I do believe that this user is well-intentioned and working to better the encyclopedia. But I would like to understand why they are adding sources that seem to come from ChatGPT, while also denying using any LLMs at all. --Gurkubondinn (talk) 22:50, 3 December 2025 (UTC)
- I think the situation has been resolved amicably, and can be closed. The user has disclosed their use of LLMs and promised not to use them in the future, and no AISIGNS have been detected other than the utm signal. Ca talk to me! 13:38, 4 December 2025 (UTC)
- Yes, very much so. I'm talking to them on User talk:Polymath Want To Be § AINB notice, and there's nothing but good intentions and a genuine interest to learn and to contribute. --Gurkubondinn (talk) 14:15, 4 December 2025 (UTC)