Unverified50% confidenceDecision RuleExact time

On a small test set of only one hundred scenarios, scores can swing by several percentage points between different runs, and judgment should be made on the full dataset.

1

Sources

50%

Confidence

Long-term

Relevance

10/12/2026

First Seen

Sources

Action Safety for AI Agents: Deciding What They're Allowed to Do

Related Entities

Test Setconcept

Related Claims

Unverified运行10次测试置信区间宽度约±31%，100次收窄至±10%，400次进一步收窄至±5%，推荐关键测试用例至少运行20-50次76% similar UnverifiedIn many evaluations conducted on small datasets, scores fluctuate by several percentage points across different runs.75% similar Unverified在5%显著性水平下测试100个随机因子，期望会有5个看起来显著（多重假设检验问题）74% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/1005161

API

curl https://kongchang.com/api/v1/knowledge/claims/1005161

MCP

get_claim(id=1005161)