Simulating Human Memory with Language Models

Qihan Wang, Nicholas Tomlin, Michael Hu +2

Language models are increasingly being deployed as user simulators, but their memory is far more reliable than that of real users. To measure this gap, we run a series of classic m…