Evaluation Backends

EvalScope supports multiple evaluation backends to integrate various tools for different evaluation tasks, detailed as follows:

OpenCompass

Initiate lightweight OpenCompass evaluation tasks through EvalScope, supporting various LLM evaluation tasks.

OpenCompass

VLMEvalKit

Initiate lightweight VLMEvalKit evaluation tasks through EvalScope, supporting various VLM evaluation tasks.

VLMEvalKit

RAGEval

Support for RAG and Multimodal RAG Evaluation: Supports independent evaluation of the retrieval module, as well as end-to-end generation evaluation.

RAGEval