Neural networks trained on synthetic task distributions to perform in-context inference without parameter updates at test time.