Unverified50% confidenceBenchmarkExact time

ItaSoRL实验中智能体的内部状态探测准确率仅约50%，等同于随机猜测水平

1

Sources

50%

Confidence

Long-term

Relevance

8/8/2026

First Seen

Sources

ItaSoRL实验揭示：AI智能体为何察觉不到世界破绽

Related Claims

Unverified一个随机猜测的模型在单次二元预测中有50%的概率押中71% similar Verified目前顶尖模型在SWE-bench Verified子集上的解决率已突破50%70% similar Unverified泰特洛克的研究得出结论：专家预测的准确率仅略高于随机猜测，甚至不如简单的统计基准线68% similar UnverifiedMedIQ多智能体系统最终达到57%准确率，摘要智能体带来约5%提升，而拥有全部前置信息的Oracle上限为66%67% similar Unverified该系统在50个已验证的CVE样本上实现了85%以上的检测准确率67% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/707452

API

curl https://kongchang.com/api/v1/knowledge/claims/707452

MCP

get_claim(id=707452)