evalscope-docs

What problem does it solve?

It helps you quickly find the right EvalScope documentation to configure and run LLM/VLM benchmark evaluations, including datasets, evaluation backends, performance stress tests, arena comparisons, and visualization.

Core Features & Use Cases

• Evaluation workflows: Learn how to run evalscope eval and evalscope perf, and how to construct TaskConfig/run_task for repeatable experiments.
• Dataset & backend guidance: Identify supported datasets/benchmarks and choose evaluation backends such as Native/OpenCompass/VLMEvalKit/RAGEval, plus advanced options like custom datasets and multi-modal evaluation.
• Advanced modes and interpretation: Use arena mode for model comparisons and visualization for inspecting results, including guidance for integrating with vLLM/Swift/SGLang.

Quick Start

Use the evalscope-docs skill to look up the correct commands and TaskConfig options for running an evaluation with your chosen model and datasets.