1 article
New framework optimizes total AI costs by accounting for inference-time scaling alongside training.