arXiv stat.ML
· Papers
Elicitation without Backpropagation: Steering Model Behavior by Optimizing the Latent Posterior
arXiv:2607.18804v1 Announce Type: cross Abstract: In the emph{latent posterior model} of transformer behavior, the next-token distribution arises from a posterior over latent predictive models conditioned on the context, mixed to generate continuations. We exploit this model in settings where it is exact, namely Bayes