Unverified50% confidenceFactExact time
The audit benchmark covers NLI (e-SNLI), adversarial syntactic shortcuts (HANS), synthetic clinical triage (N=900), prompt injection defense (Tensor Trust, N=250), and an 86-case logic-gate trigger suite.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe study designed a utility scoring system: correct answers gain points, incorrect answers lose points, and abstaining is a neutral safe option61% similarUnverifiedThis counterfactual audit covers seven open-weight models (8B to 70B parameter scale) and four benchmarks spanning judicial reasoning and contract reasoning.60% similarVerifiedUnder healthcare safety requirements, the industry generally recommends a dual-verification mechanism of AI pre-screening plus manual confirmation for prescription parsing.60% similarUnverified针对提示词注入的业界推荐多层防御体系包括输入沙箱化、输出格式校验和关键操作人工审批60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1155158API
curl https://kongchang.com/api/v1/knowledge/claims/1155158MCP
get_claim(id=1155158)