Exponential gates and matrix memory: xLSTM and RWKV-7
Exponential gating vs sigmoid; matrix-valued memory; vector-valued gating; in-context learning rates.
Exponential gates and matrix memory: xLSTM and RWKV-7
This chapter is planned. See the chapter-at-a-glance (foundations §3) for the full design; the crosswalk (foundations §22) maps this chapter to the relevant post_transformers/roadmap.md weeks.