[BUG] Claude LLM - Improper use of verb vs noun hybrid words

Environment

• Platform (select one):
• [x] Anthropic API
• [ ] AWS Bedrock
• [ ] Google Vertex AI
• [x] Other: Claude LLM itself?
• Claude CLI version: NA (LLM training issue)
• Operating System: NA
• Terminal: NA

Bug Description

Set up is a verb. Setup is a noun.
Log in is a verb. Login is a noun.
Clean up is a verb. Cleanup is a noun.

Claude pretty consistently always uses the noun form. If you're performing the action, it is always two words. If you're talking about it as a thing, one word. I use my racecar to race cars. I definitely don't use my race car to racecars.

These are all incorrect, and are methods that Claude output for me in the span of a day:
SetupUserInfo() (should be SetUpUserInfo())
LoginUserAsAdmin() (Should be LogInUserAsAdmin())
CleanupOrphanedRecords.sql (Should be CleanUpOrphanedRecords.sql)

These are also incorrect (but are not mistakes Claude currently has made for me - but if making training data, wouldn't want it to blindly overcorrect the other direction):
GetUserInfoSetUps() (noun, should be Setup -- get is the verb, Setup is the thing).
GetUserLogIns() (a Login page, or a user Login, is used to LogInToSomething. These are User Logins which the user would use to log in).

Steps to Reproduce

1. Ask Claude to create a cleanup script. Ask Claude to create a setup method. Ask Claude to create a login method.

Expected Behavior

Shouldn't output confusing / grammatically incorrect casing, harming readability and perpetuating improper use of common programming terms.

Actual Behavior

Casing all over the place for hybrid words - mostly leaning towards the noun hybridization with no regard to whether it is an action or a subject.

Additional Context

This is a very common programmer mistake so I'm sure its just bad training data. Even my bank has a button titled "Login" on their login page.
Nitpicky, but combined with Git being obnoxious about changing file name casing on windows it gets really annoying to fix a bunch of these if you don't notice it before the initial commit. Seems real easy to fix at the training level, I'd imagine.
I couldn't find a better place to file this bug, feel free to point me in the right direction if there's an official spot to report non-security LLM training issues.