Talk:KMV credit correlation model

Comments from AINB reviewer

I have checked up to partway through "The correlation problem" only.

In (1) an author is given as Claude Lanfranconi but this person is Mariano Lanfranconi in every mention I can find.

The statement about founding is not confirmed by what I can see in (5) but it's possible a subscriber sees more. In general (5) is a terrible source. The information is present in (6) but not cited to that. The article correctly notes that (5) and (6) disagree on the founding date but claims "some accounts" in plural while only showing one (6) (a very typical recent-LLM issue).

the founders' own recollections place the origins of the the origin of the venture in the late 1980's, following earlier collaboration on a loan-pooling business dating to 1986 is supported by neither (6) nor (7).

As a larger issue, why does this entire History section exist? It seems to be a history of the KMV company, but that is not the topic of the article and the vast majority of the information here appears irrelevant. It may be relevant to describe how KMV came to develop this methodology, but surely details of how and when it was purchased are off topic.

KMV's principal products were Credit Monitor, which produced Expected Default Frequency (EDF) estimates for individual firms, and Portfolio Manager, which applied the correlation model at portfolio level. The correlation model itself was first released commercially in 1996. is sourced to 8, which has nothing to say about the first sentence and does not really confirm the second. The first sentence seems WP:UNDUE as this article is not about KMV company.

The equation in "The Merton framework" has terms not defined here. I examined the Wikilinked articles but they put forward extremely different formulations which I can't relate to this one. Without a definition of N() I can't check the correctness of the equation.

The source given as Crosbie, Peter; Bohn, Jeffrey (2003). Modeling Default Risk (Report). Moody's KMV has been published as [] which says it's from Moody's MVK (not KMV) "White Paper" series. Prefer published version.

So far I'm seeing two concerns. One is a pattern of small source/text inconsistencies where the information may be true, but the given source does not confirm it. This is absolutely typical of 2026 LLM sourcing. The other is topic drift, where material is introduced that does not relate to the article topic. This is also highly typical of LLM. Both are errors that humans can make as well, but they're pervasive enough here to make me feel that LLM may have been used for more than the stated copyediting. M kuhner (talk) 18:34, 8 August 2026 (UTC)