Unverified50% confidenceBenchmarkExact time

在446次两两对比中，CTD在21.1%的情况下表现相当，仅在6.1%的情况下统计显著劣于对手

1

Sources

50%

Confidence

Long-term

Relevance

9/15/2026

First Seen

Sources

先收敛后多样化：多目标贝叶斯优化的CTD新范式

Related Entities

CTD (Converge-Then-Diversify)concept

Related Claims

Unverified在总计446次两两对比中，CTD在72.9%的情况下统计显著优于当前最先进方法77% similar UnverifiedDeepSuite基准测试的误报率为44.9%，漏报率为1.2%64% similar Unverified在奖励差距较大的智能体配对上，决策分歧率低于1%64% similar Unverified在多轮叙述作用下，模型的最终判断相比匹配的单轮基线平均偏移了25个百分点64% similar Unverified在MATH数据集的Level 5竞赛级难题上，前沿模型准确率仍低于60%63% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/921697

API

curl https://kongchang.com/api/v1/knowledge/claims/921697

MCP

get_claim(id=921697)