CI-1T FAQs

Question 1

Can I use CI-1T to evaluate my own LLMs without consuming credits?

Accepted Answer

Yes, CI-1T features BYOM (Bring Your Own Model) probing. This allows you to evaluate the instability of local or remote OpenAI-compatible LLMs directly, without requiring CI-1T credits or an API key for this specific mode.

Question 2

Does CI-1T include any local utility tools?

Accepted Answer

Yes, CI-1T provides several local utility tools that work without an API key or consuming credits. These include tools for score interpretation, conversion (e.g., to Q0.16), configuration generation, drift detection, custom alerting, and interactive visualization of evaluation results.

Question 3

How does CI-1T help evaluate AI model stability?

Accepted Answer

It provides comprehensive stability metrics like Collapse Index (CI) and Authority Level (AL), supports fleet-wide multi-node evaluation, and offers local tools for drift detection, custom alerting, and interactive visualization of prediction stability results.

Question 4

What are the benefits of CI-1T's fleet session management?

Accepted Answer

Fleet session management enables persistent multi-node evaluation across distributed environments. You can create, manage, and monitor sessions for multiple models or nodes, streamlining assessment and ensuring consistent stability monitoring across your AI infrastructure.

Question 5

What is CI-1T and its primary function?

Accepted Answer

CI-1T is an MCP (Model Context Protocol) server that allows AI agents to evaluate the stability of prediction models, manage distributed fleet sessions for multi-node assessment, and directly control API keys for the CI-1T prediction engine.

CI-1T

CI-1T

主な機能

ユースケース

主な機能

ユースケース