Unverified50% confidenceBenchmarkExact time
Native decision models achieve 62.50% accuracy on the TandemBench benchmark, but are close to random on optimality verification.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/5/2026
First Seen
Valid until: 2/3/2027
Sources
Related Entities
Related Claims
UnverifiedWhen asked directly, the accuracy of the decision models falls in the 50.0%–67.0% range.77% similarUnverifiedIn the ItaSoRL experiment, the accuracy of probing the agent's internal states was only approximately 50%, equivalent to random guessing.74% similarUnverified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73% similarUnverifiedA linear probe classifying the selection arm achieves an accuracy of up to 59.7% (chance level 50%), indicating that the selection behavior leaves detectable traces in the model's representations.73% similarUnverifiedOpportunity-normalized accuracy is computed as (model accuracy − random-guess accuracy) / (1 − random-guess accuracy).72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1137706API
curl https://kongchang.com/api/v1/knowledge/claims/1137706MCP
get_claim(id=1137706)