The Phi-3 paper introduces highly performant iPhone LLMs, with different models and training techniques, such as data optimization and post-training fine-tuning. Pairing Phi-3 with a search engine can enhance its performance on tasks requiring wide scopes of knowledge.
Table of contents
Key TerminologyPhi 3 ArchitectureTraining and Data Optimal MixesPost-TrainingQuantization for Phone UsagePairing Phi-3 with SearchConclusionSort: