Welcome to the OpenCompass Documentation¶
OpenCompass is a one-stop evaluation platform for large language models and multimodal models. The documentation is organized so that you first complete an evaluation, then understand and extend each component.
If this is your first time using OpenCompass, read Installation and Environment Setup, Five-Minute Quick Start, and Running a Complete Evaluation from a Configuration in order. If you already have a model or dataset, you can go directly to the corresponding topic.
Get Started
Basic Tutorials
Prompts and Input Construction
Execution, Parallelism, and Task Management
Evaluation Methods
Extending OpenCompass
Tools and Recovery