To address this, we approach the problem from the perspective of evolving question complexity and select problems from Code Evol-Instruct dataset [10], which evolves in depth and breadth to cover a wide range of complexities and has proven effective in non-reasoning models [11].
Hello author, I am wondering how the dataset be curated. How to generate short and long reasoning data?
To address this, we approach the problem from the perspective of evolving question complexity and select problems from Code Evol-Instruct dataset [10], which evolves in depth and breadth to cover a wide range of complexities and has proven effective in non-reasoning models [11].Hello author, I am wondering how the dataset be curated. How to generate short and long reasoning data?