Long Context Coding Evaluation Dataset for Gemini CLI

This project builds a long context and complex reasoning coding evaluation dataset for Gemini CLI. It focuses on multi file, multi language software engineering tasks, difficulty certification, agentic evaluation, and structured failure analysis to better measure real coding agent performance on large repositories.