I tried the same test (iPad Pro 4. Gen, iPadOS 15 Beta2, Xcode 13 Beta 2) with my AR app and the motion capture sample app from Apple. But in the logs I observe errors and the character does not become visible. I wonder how you were able to run your motion capture app and get the character visible. The motion capture improvements were announced in WWDC21 for devices with A14 chips. Maybe they are not enabled for devices with A12 and A13 chips.

I can confirm the problem. I use an iPad Pro 4. Generation with iPadOS 15 Beta2, Xcode 13 Beta 2 and build the motion capture sample. When I debug the app I observe the robot character is loaded and attached to the anchor, but the robot is not visible. There are some errors in the log: [Builder] Unable to create skeletal pose mesh asset at path (.../ace_PLY) because a model descriptor wasn't successfully created. throwing -10878 [Technique] ARWorldTrackingTechnique: World tracking performance is being affected by resource constraints [Builder] Skeletal pose asset does not exist. Cannot setup the skeletal pose component. Obviously Motion Capture is buggy and not working on iOS/iPadOS 15 Beta 2.

It is not even possible to test the motion capture improvements since with iOS 15 Beta 2 motion capture does not work at all. Refer to this entry

Problem still reproducible with iPadOS 15 Beta3, Xcode 13 Beta 3.