用於評估本地LLM品質的社群基準測試套件。透過API提交結果。
A five-prompt creative writing eval where models draft short tech-related 4chan-style greenposts. DeepSeek judges format compliance, reasonable length, tech relevance, coherence, and humor.