Unverified85% confidenceConceptTime unknown
大语言模型幻觉（Hallucination）在多轮对话或多步推理的Agent场景中危害被成倍放大，形成「幻觉级联」（Hallucination Cascade）
1
Sources
85%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Agent上下文腐烂的两种病因：Distraction与Poisoning详解
bilibiliAI大模型原理
Related Entities
Related Claims
UnverifiedTraditional end-to-end LLM calls often suffer from hallucinations or logical leaps when facing multi-step, multi-constraint complex tasks.72% similarUnverified大型语言模型在处理需要长程上下文精确追踪的细节一致性问题时容易产生幻觉性漂移（Hallucination Drift）71% similarUnverified模型规模越大，幻觉的表述往往越流畅、越具迷惑性，识别难度反而上升70% similarVerified在多Agent场景中，幻觉问题会被级联放大：一个Agent的错误输出可能成为下一个Agent的输入，导致错误在链路中不断累积69% similarUnverifiedHallucination refers to LLMs generating plausible-sounding but factually incorrect information69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/22390API
curl https://kongchang.com/api/v1/knowledge/claims/22390MCP
get_claim(id=22390)