Unverified50% confidenceSolutionExact time

Critic agent 逐条审视 Detector 提出的 pattern 并反驳站不住脚的结论，最终只保留三到四条最扎实的

1

Sources

50%

Confidence

Medium-term (~90 days)

Relevance

7/22/2026

First Seen

Valid until: 10/20/2026

Sources

Teaching AI to Have "Taste": Exploring Multi-Agent Design with Open-Source Skill Taste

Related Claims

Unverified对边缘分数论文，建议完整走完rebuttal观察分数走向，若分数无实质提升且无缘主会与Findings则撤稿转投对口workshop62% similar UnverifiedAgent 评测中由于错误沿任务链累积传播，最终完成率的分差往往比传统 NLP 基准更难弥合60% similar Unverified对抗测试的效果高度依赖场景库的完备性，未被预见的攻击路径依然是盲区，单次通过测试不代表Agent在生产中永远安全59% similar Unverified在多步骤 Agent 工作流中，未被拦截的脏数据会触发下游节点连锁失败58% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/593912

API

curl https://kongchang.com/api/v1/knowledge/claims/593912

MCP

get_claim(id=593912)