Unverified50% confidenceBenchmarkExact time
In the prompt-injection defense task (N=250), System 1 models with 4B parameters or more achieve an AUROC of 0.988 to 0.998, a recall of up to 95.2%, and a 0.0% false positive rate at 48 to 60 milliseconds of latency.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedUsing Jensen-Shannon divergence between time-domain and frequency-domain models as an uncertainty measure, AURC was 0.29 on the standard split (better than 0.36) and 0.54 under single-condition reproduction (better than 0.73).70% similarUnverifiedUsing the behavioral features of each state for failure prediction, the AUROC on the held-out set reaches as high as 0.94.70% similarUnverifiedReachNT's urgency model achieves a ROC-AUC of 0.84 on unseen phrasing, and as part of an overall system that includes human review it catches 99.3% of dangerous work orders.70% similarUnverified当训练窗口、门控窗口和报告窗口未严格分离时,PR-AUC达到0.72到0.76,严格分离后骤降至约0.51接近随机水平,这是数据泄漏的表现69% similarUnverifiedOn the hard tier, recall@5 improved from 0.000 to 0.955 (under default blending weights), and reached 1.000 when using the lower-bound variant.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1155170API
curl https://kongchang.com/api/v1/knowledge/claims/1155170MCP
get_claim(id=1155170)