#research

Stanford's framework for programming, not prompting, language models, with optimizers that tune your prompts for you.

A Microsoft Research library that compresses long prompts by dropping low-value tokens, so the model reads less and costs less.