First-order ANIL provably learns representations despite overparametrisation

Published in ICLR 2024, 2023