Atlas

Benchmarks

← All benchmarks

CCPM

General QA · 2021-06-03

CCPM (Chinese Classical Poetry Matching) gives a modern-Chinese description and asks which of four candidate lines of classical Chinese poetry it describes. Scores are accuracy. Higher is better.

Top models (higher is better)

ModelScore
DeepSeek-V3-Base92.0
DeepSeek LLM 67B Base88.5
Qwen2.5 72B88.5
Qwen1.5 72B88.1
DeepSeek LLM 67B Chat84.9
Llama 3.1 405B78.6
DeepSeek LLM 7B Base76.9
DeepSeek LLM 7B Chat76.5
Llama 3 70B68.1
Mixtral 8x22B61.0
Loading Atlas data…