Cross-Model Memory Transfer via Target-Side Reader Adaptation

Summary

Methods for improving knowledge use in LLMs typically fall into non-parametric retrieval or parametric adaptation. Engram-style hashed memory stores learned information in an external, addressable table, consumed through a small learned reader. This study investigates what matters more when such memory is transferred across backbones: the frozen memory or the target-side reader, via cross-model frozen-memory extraction. Ablations show that learned memory content and correct addressing both matter, but the transferred table becomes useful only through a reader aligned to the target model. In downstream QA tasks, a dual-layer, four-branch reader nearly closes the same-model versus cross-model reuse gap, achieving an average score of 38.8. When the provider reader is directly compatible, the frozen artifact provides utility without target-side training, and optional adaptation yields further improvement. These results suggest Engram can serve as a reusable external knowledge artifact with a compatible reader interface.