Unverified50% confidenceDecision RuleExact time
When testing prompt injection, change only one variable at a time, and since model behavior is probabilistic, tests should be repeated multiple times and results reported as a rate (for example, three successes out of ten).
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
Unverified对同一测试用例执行多次调用统计成功比例得到通过率,比二元的通过/失败判断更能反映模型真实表现78% similarUnverified多重比较问题指同时测试多个模型变体时,即使每个单独检验显著性水平设为0.05,整体假阳性率也会急剧膨胀76% similarUnverifiedWhen testing LLM citation behavior, the same prompt should be run at least 20–50 times, ideally expanded to more than 100 times.75% similarUnverified在机器学习多种子实验中重复次数通常只有5到10次,样本量极小,正态性假设难以验证,因此 Wilcoxon 检验往往是更保守稳健的选择75% similarUnverifiedRetry rate refers to the proportion of cases where the model's first output is not adopted and the developer needs to ask again or adjust the prompt and try again; it is ubiquitous in real workflows because the model's one-shot success rate is far below the ideal value assumed by benchmarks.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1066752API
curl https://kongchang.com/api/v1/knowledge/claims/1066752MCP
get_claim(id=1066752)