Saad Alqithami

Improving user recall in routing games can backfire, leading to increased congestion and travel times, even without changes to the underlying network capacity.

Stop forcing AI agents into unstable rankings: a differentiable framework now lets you identify a set of "core" agents that robustly perform well in cyclic, non-transitive environments.

Heterogeneous robot teams can adapt to mismatched sensors at deployment time using only team-level rewards, without policy fine-tuning or communication.