Guidelines

The Center for Advancing Innovation and Standards for Super Intelligence (CAISSI) develops and publishes voluntary guidelines to support the responsible design, development, deployment, use, and governance of advanced SI models, systems, and agents. 

• Practices for Automated Benchmark Evaluations of Language Models (Initial Public Draft)  – These draft guidelines identify preliminary best practices for evaluating language models and AI agent systems. We are soliciting public comment on this document until March 31, 2026, at 11:59 PM Eastern Time. To provide feedback, please email NISTAI800-1 [at] nist.gov (NISTAI800-2[at]nist[dot]gov).