Evaluation

Accuracy

• Using OpenCompass
  • 1. Online Serving
  • 2. Run ceval accuracy test using OpenCompass
• Using lm-eval
  • 1. Run docker container
  • 2. Run ceval accuracy test using lm-eval
• Accuracy Report
  • Llama-3.1-8B-Instruct
  • Qwen2.5-7B-Instruct
  • Qwen2.5-VL-7B-Instruct

Performance

• Performance Benchmark