# Cross-Model Transfer of Recurrent Memory

In one architecture-matched Qwen3.5 4B→9B hybrid (Gated DeltaNet plus full attention) sibling pair, the work installs the source model's persistent inference state after reading a prefix — attention KV plus GDN recurrent matrices and convolution history — directly into the larger receiver with no target prefix replay; holding translated KV fixed, adding the GDN persistent-state package lowers teacher-forced NLL by 0.747 nats/token (95% paired document bootstrap CI [0.6921, 0.8047]) and improves all 64 PG19 documents, direct recurrent and convolution reuse beats the tested learned GDN maps, and a rank-4 correction with 434,176 trainable parameters brings continuation loss to only 0.076 nats/token above native 9B, JS divergence 0.022, native context recovery 0.

In one architecture-matched Qwen3.5 4B→9B hybrid (Gated DeltaNet plus full attention) sibling pair, the work installs the source model's persistent inference state after reading a prefix — attention KV plus GDN recurrent matrices and convolution history — directly into the larger receiver with no target prefix replay; holding translated KV fixed, adding the GDN persistent-state package lowers teacher-forced NLL by 0.747 nats/token (95% paired document bootstrap CI [0.6921, 0.8047]) and improves all 64 PG19 documents, direct recurrent and convolution reuse beats the tested learned GDN maps, and a rank-4 correction with 434,176 trainable parameters brings continuation loss to only 0.076 nats/token above native 9B, JS divergence 0.022, native context recovery 0.