Unverified50% confidenceFactExact time

顺序写入的性能通常比随机写入高出一到两个数量级

1

Sources

50%

Confidence

Long-term

Relevance

8/25/2026

First Seen

Sources

Apache Maka: An Auditable, Local-First AI Agent Workspace

Related Entities

追加式日志concept

Related Claims

UnverifiedIn few-shot prompting, simply changing the order of examples can cause dramatic fluctuations in accuracy, ranging from near-random to near-optimal.75% similar Unverified在某些任务上，仅因随机种子不同导致的性能波动幅度可能超过某些论文声称的方法创新带来的提升71% similar Unverified权重初始化的随机性对模型性能影响最深远，数据批次打乱顺序的随机性影响最小69% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/798587

API

curl https://kongchang.com/api/v1/knowledge/claims/798587

MCP

get_claim(id=798587)