Question 1

What is Scanf?

Accepted Answer

Scanf is Telluvian's real-time AI hallucination detection tool. It analyses a model's internal activations as it generates a response and returns a confidence signal alongside every output, so you know which parts of an answer to trust.

Question 2

How does Scanf detect hallucinations?

Accepted Answer

A query is submitted to a frontier model. While the response is generated, Scanf's probing classifiers inspect the model's internal activations in real time. The response is returned together with a token-level trust indicator showing where the model was uncertain.

Question 3

Does Scanf need access to a model's weights?

Accepted Answer

Not always. For open-weight models, Telluvian inspects activations directly. For closed models, Telluvian uses an approach called Proxy White-Box Interpretability: measuring directions like refusal or uncertainty in an open-weight model with a similar training distribution, and using that as an implicit proxy measurement inside the closed model.

Question 4

Who is Telluvian for?

Accepted Answer

Legal teams, who use Scanf to catch hallucinated case citations and misquoted statutes in AI-assisted drafting; finance and compliance teams, who verify AI outputs in regulated workflows; and AI product teams, who integrate Scanf as a verification layer to give their own users confidence signals on every output.

Question 5

What is mechanistic interpretability?

Accepted Answer

Mechanistic interpretability is the practice of understanding what's happening inside a language model, rather than treating it as a black box, by studying its internal activations. Telluvian uses mechanistic interpretability techniques to build the probing classifiers behind Scanf.

Your portal to AI