Common Unified Benchmark Environments — a protocol standard that eliminates the integration tax of agentic benchmarks by providing a universal interface between benchmarks and evaluation frameworks.
ArXiv Paper — Read the paper detailing the vision for this project.
About CUBE — learn more about the project, the problem it solves, and who it’s for.
Get Started
- Authoring a CUBE — How to wrap a benchmark as a CUBE: three starting paths, implementation order, validation, and submission
- Design Philosophy — Read before proposing a change to the framework itself: the broader picture and the bar a change must clear
- GitHub README — Installation, quick start, and contributing
- DeepWiki — Full API reference and guides