According to StanfordAILab, Training Witnesses offers verifiable “honesty” certificates to ease ML conference submission overload and rebuild trust.