KodCode-V1 Encoding Synthetic Dataset

KodCode was released in 2025 by researchers from Microsoft GenAI, the University of Washington, and the University of Texas at Austin.KodCode: A Diverse, Challenging, and Verifiable Synthetic Dataset for Coding".

The dataset is the largest fully synthetic open-source dataset that provides verifiable solutions and tests for coding tasks. It contains 12 different subsets covering various fields (from algorithms to package-specific knowledge) and difficulty levels (from basic coding exercises to interviews and competitive programming challenges), and is designed for supervised fine-tuning (SFT) and RL tuning.