Technical FAQ
How do you prevent hallucinations?
Our engine uses a deterministic reasoning layer that requires every output to be synthesized from a verifiable chain of vector relationships. If a logic path cannot be validated against the source data, the system flags it as Inconclusive rather than guessing.
What is the latency of the reasoning API?
Our core engine is optimized for high-frequency environments. Standard reasoning synthesis for complex financial vectors averages 12–15ms. Custom deep-vector training models may vary based on dataset complexity.
How is data vectorized for cybersecurity?
We map network behaviors, adversary tactics (MITRE ATT&CK), and system logs into a multi-dimensional vector space. This allows our engine to reason across disparate data sources to identify the logical progression of a threat.
Is my proprietary data used to train global models?
No. Interdot employs strict data isolation protocols. Custom vector training is performed in siloed environments, and your data never leaks into our baseline reasoning engine.
Can I audit the reasoning chains?
Yes. Every API response includes a Logic Trace object — a step-by-step breakdown of the causal relationships the engine used to reach its conclusion.