DeepSeek’s Own Paper Says Few-Shot Prompting “Consistently Degrades” Its Reasoning Model
中文摘要
DeepSeek论文指出，少样本提示会持续削弱其推理模型的表现，提示词优化可能已不再适用。
English Summary
DeepSeek's paper reveals that few-shot prompting consistently degrades its reasoning model's performance, suggesting previous prompt optimizations are now obsolete.
Original Excerpt
Your prompt was tuned against a model that no longer exists Continue reading on Medium »