Unverified50% confidenceFactExact time
语义等价的不同表述可能导致模型回答质量出现高达30%的波动，该现象被称为'提示敏感性'（prompt sensitivity）
1
Sources
50%
Confidence
Long-term
Relevance
8/26/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIncluding specific evaluation dimensions in prompts (such as format, word count, key points to cover, and items to exclude) can improve model output quality by an average of 30%–50% compared to vague prompts.71% similarUnverified同一个基座模型在不同系统提示词下的表现差异可以高达30-50个百分点66% similarUnverifiedFew-shot提示通过提供2-5个示例让模型理解模式，比纯文字描述准确率通常高出20%-40%64% similarUnverified2024 年的多项基准测试表明，经过严格安全对齐的模型在代码生成任务上的拒绝率比基座模型高出 15-30%，在创意写作任务上的自我审查率高达 40%64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/802467API
curl https://kongchang.com/api/v1/knowledge/claims/802467MCP
get_claim(id=802467)