Runs an AI belief challenge endpoint to test or compare model beliefs.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.