Unverified50% confidenceFactExact time
The CoM method was evaluated on elicitation tasks for sensitive topics such as CBRN materials, explosives and illegal drugs, and was compared with the earlier preference-boundary (PRBO) method and a direct-training baseline.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/30/2026
First Seen
Valid until: 1/28/2027
Sources
Related Entities
Related Claims
UnverifiedClaude's application in drug R&D may concentrate on stages such as target research, auxiliary analysis for compound screening, and clinical trial protocol design.64% similarUnverifiedThe laboratory adopted a phased validation strategy: first verifying through small-scale training and evaluation whether newly deployed safety measures are effective, and building more evidence regarding alignment.63% similarUnverifiedCommon practices for aligning high-risk capabilities include refusal fine-tuning in the post-training stage for tasks such as vulnerability discovery, malicious code generation and penetration testing, or interception via system prompts and classifiers.63% similarUnverifiedThe chemical reaction prediction task is used as a proxy metric for CBRN (chemical, biological, radiological, nuclear) related hazardous capabilities.63% similarUnverified应在验证指标达到峰值时保存模型检查点以避免后续训练破坏已学策略，这是最务实的工程防线62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1114194API
curl https://kongchang.com/api/v1/knowledge/claims/1114194MCP
get_claim(id=1114194)