Your 100K-Token Prompt Is Making the Model Dumber. All 18 Frontier Models Failed the Same Way.
中文摘要
Chroma 测试显示，所有 18 种主流 AI 模型在输入增加时性能均会下降，超长提示词会削弱模型智力。
English Summary
Chroma's testing revealed that all 18 frontier AI models performed worse as input length increased, showing that long prompts consistently degrade model performance.
Original Excerpt
Chroma tested 18 frontier models and every single one got worse as input grew. Continue reading on Medium »