Mark Dredze

LLM agents can achieve high task accuracy while exhibiting profound epistemic *im*humility, failing to acknowledge uncertainty even when presented with contradictory evidence.

Weird generalization is not just a quirk of model training; it’s a fragile phenomenon that could be weaponized if not carefully managed.

Myopic planners can fail spectacularly when faced with the need to acquire capabilities for future experiments, leading to unbounded approximation ratios in goal-directed discovery.