On the Robustness of Langevin Dynamics to Score Function Error

We consider the robustness of score-based generative modeling to errors in the estimate of the score function. In particular, we show that Langevin dynamics is not robust to the $L^2$ errors (more generally $L^p$ errors) in the estimate of the score function. It is well-established that with small $L^2$ errors in the estimate of the score function, diffusion models can sample faithfully from the target distribution under fairly mild regularity assumptions in a polynomial time horizon. In contrast, our work shows that even for simple distributions in high dimensions, Langevin dynamics run for any polynomial time horizon will produce a distribution far from the target distribution in Total Variation (TV) distance, even when the $L^2$ error (more generally $L^p$) of the estimate of the score function is arbitrarily small. Considering such an error in the estimate of the score function is unavoidable in practice when learning the score function from data, our results provide further justification for diffusion models over Langevin dynamics and serve to caution against the use of Langevin dynamics with estimated scores.

We study the effects of errors in the "score function" on the behavior of generative AI algorithms. We mathematically show that even in simple cases, a popular, simple method known as Langevin Dynamics is not robust to such errors. This lack of robustness causes the outputs generated by the algorithm to be very different than the desired ones. As these errors are inevitable in practice, our result suggests Langevin Dynamics is not an effective method for generative AI. Our result contrasts with the popular method of diffusion models, which have been established to be robust to such errors.