Unverified50% confidenceTradeoffExact time
When multiple Agents form a collaborative network, 'hallucination contagion' may occur, where erroneous information generated by one Agent is accepted as fact by another Agent and further reasoned upon, causing errors to propagate through the network.
1
Sources
50%
Confidence
Long-term
Relevance
9/8/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedHallucinations in agents directly translate into erroneous actions, including calling non-existent tools, fabricating file contents, and inventing error messages and then executing subsequent operations based on them.70% similarUnverifiedAgent链式调用多个技能时，若某个技能被恶意构造或存在漏洞，可能引发提示注入攻击（Prompt Injection）66% similarUnverifiedAn agent's behavioral chain is essentially a probabilistic reasoning process in which the same input may produce different tool-call sequences, which is what makes agent evaluation difficult.66% similarUnverified多Agent循环协作模式中，生成Agent产出初稿、批评Agent指出问题、修订Agent整合反馈，三者形成多轮迭代协作循环63% similarVerifiedThe principle of indirect prompt injection is that an attacker disguises malicious instructions as normal data, delivers them to the model through a data source trusted by the agent, and induces the model to interpret the data as instructions and execute them.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/877468API
curl https://kongchang.com/api/v1/knowledge/claims/877468MCP
get_claim(id=877468)