Training scenario where ground-truth success signals are unavailable and must be inferred from other feedback.