Unverified50% confidenceBenchmarkExact time
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%
1
Sources
50%
Confidence
Long-term
Relevance
7/18/2026
First Seen
Sources
Related Claims
UnverifiedIn evaluations of early Agent systems, a Stanford research team found that a single-step error rate of 5% can lead to a final failure rate exceeding 40% in 10-step tasks.73% 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71% similarUnverifiedEven GPT-4-level Agents often achieve end-to-end success rates below 60% on complex tasks involving more than 5 tool calls.70% similarUnverified斯坦福大学2023年的研究显示，当时主流LLM对间接提示注入的防御成功率普遍低于30%69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/552253API
curl https://kongchang.com/api/v1/knowledge/claims/552253MCP
get_claim(id=552253)