Search papers, labs, and topics across Lattice.
1
0
3
User simulators trained on existing datasets may be optimizing for unrealistic scenarios, as a new benchmark reveals a significant "realism gap" in their ability to generalize to diverse recommender behaviors.
