Anay Mehrotra

The paper introduces Entropy-Cut Metropolis-Hastings, an efficient sampling method that uses next-token entropy to identify and resample from critical decision points in a reasoning trace, significantly improving sampling efficiency over existing uniform cut methods.

The paper analyzes language generation and identification in the limit under bounded memory, showing that memory constraints significantly alter learnability, particularly affecting achievable density and convergence guarantees.

The paper proposes a novel, computationally efficient estimator for estimating heterogeneous treatment effects in panel data by framing the problem as matrix completion and establishing a sharp row-wise $\ell_2$ error bound.