Apple Intelligence Foundation Language Models [pdf]

[azhenley]

https://machinelearning.apple.com/papers/apple_intelligence_foundation_language_models.pdf

[dagmx]

This is a pretty fantastic , high level overview of their system.

I also appreciate that they cover infrastructure and data collection aspects, which are often glossed over.

[elpakal]

> Code data is obtained from license-filtered1 open source repositories on GitHub. The bulk of the code data covers 14 common programming languages, including: Swift, Python, C, Objective-C, C++, JavaScript, Java, and Go.

That's interesting, looking forward to seeing how accurate the output is for Swift.

[ActVen]

The on device benchmarks are impressive. Especially tool use.