S. Ulukus

Chained RLMs can boost accuracy in multi-iteration reasoning tasks by effectively managing context and correcting errors through a novel inference architecture.

A novel metric reveals that existing methods for measuring privacy leakage in multidimensional data are fundamentally limited, while the Dependency Triad offers a scalable solution that maintains accuracy even in high-cardinality scenarios.

Tailoring preemption strategies can slash average age of information by optimizing when new updates interrupt ongoing ones.

Long-time behaviors in deep linear transformers can reveal surprising dynamics like clustering and oscillations, all rooted in a hidden Hamiltonian structure.

The All-out Attack strategy reveals that miners can exploit Pay-Per-Share schemes to maximize gains while leaving honest miners and pool operators at a disadvantage.

A simple tweak to a known blockchain attack—limiting the withholding time—yields 22x greater rewards for attackers in realistic scenarios, exposing a major weakness in pooled mining incentives.

Sharing randomness across all servers in a graph-based database system boosts private information retrieval capacity, outperforming schemes where randomness is tied to individual message replicas.