mechanistic interpretability

Eliciting Nmuerical Predictive Distributions of LLMs without Autoregression
We train a set of simple probes to recover the predictive distributions of LLMs, with uncertainty estimates.
Julianna Piskorz, Kasia Kobalczyk, Mihaela van der Schaar
Jan 30, 2026
openreview
Eliciting Nmuerical Predictive Distributions of LLMs without Autoregression