Pinocchio

Pinocchio is a research preview addressing LLM hallucinations through verifiable AI outputs. Users provide tasks and source material; the system carries numeric values directly from sources rather than regenerating them, recording all computations and derivations applied.

A deterministic verifier can replay and check the complete computation chain, providing transparency into how results were derived. This approach aims to create auditable AI work with traceable reasoning.