Phi-4 Technical Report

TL;DR

Phi-4 is a large language model that emphasizes data quality and synthetic data integration, surpassing its teacher model in STEM reasoning tasks through innovative training and data strategies.

Contribution

The paper introduces phi-4, a 14-billion parameter model that improves upon previous models by focusing on data quality and synthetic data, achieving superior reasoning capabilities.

Findings

Outperforms GPT-4 on STEM QA tasks

Achieves strong reasoning performance relative to size

Utilizes innovative data-generation and post-training techniques

Abstract

We present phi-4, a 14-billion parameter language model developed with a training recipe that is centrally focused on data quality. Unlike most language models, where pre-training is based primarily on organic data sources such as web content or code, phi-4 strategically incorporates synthetic data throughout the training process. While previous models in the Phi family largely distill the capabilities of a teacher model (specifically GPT-4), phi-4 substantially surpasses its teacher model on STEM-focused QA capabilities, giving evidence that our data-generation and post-training techniques go beyond distillation. Despite minimal changes to the phi-3 architecture, phi-4 achieves strong performance relative to its size -- especially on reasoning-focused benchmarks -- due to improved data, training curriculum, and innovations in the post-training scheme.