Unverified50% confidenceFactExact time
UC Berkeley analyzed more than 1,600 real execution traces and found that nearly 40% of failures were not due to insufficient capability of a single Agent, but to information not being aligned between Agents.
1
Sources
50%
Confidence
Long-term
Relevance
10/27/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedUC Berkeley's research on multi-agent system failure taxonomy (MAST) shows that production-grade multi-agent workflows have failure rates as high as 41% to 86.7%.70% similarVerifiedAgents suffer from error accumulation in multi-step tasks; a ten-step sequential process where each step has 95% accuracy results in an overall success rate that drops to approximately 60%.69% similarUnverified2023年斯坦福研究显示，在需要连续多步决策的任务中，现有Agent系统的成功率随步骤数呈指数级下降69% similarUnverifiedAnthropic recommends starting agent evaluation with 20 to 50 tasks drawn from real failures.67% similarUnverified斯坦福大学的AgentBench基准测试（2023年）量化证明，在复杂多步骤任务中纯ReAct范式的成功率往往不超过30%66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1101844API
curl https://kongchang.com/api/v1/knowledge/claims/1101844MCP
get_claim(id=1101844)