Unverified50% confidenceFactExact time
困惑度（Perplexity）、BLEU分数、ROUGE分数等自动化指标最初为机器翻译等任务设计，在衡量创意性方面存在天然缺陷
1
Sources
50%
Confidence
Long-term
Relevance
8/19/2026
First Seen
Sources
Related Claims
UnverifiedAutomated evaluation metrics such as BLEU, ROUGE, and perplexity primarily measure surface-level statistical features and are difficult to capture issues of linguistic naturalness.73% similarUnverifiedThe perplexity of LLM output is typically lower than that of human writing, and can be used to detect AI-generated content.67% similarUnverified提示工程已从早期的关键词拼接演进为包含角色设定、思维链、少样本示例等技术的完整体系62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/772774API
curl https://kongchang.com/api/v1/knowledge/claims/772774MCP
get_claim(id=772774)