Larimar shows an encoder writing facts to external memory that conditions a frozen decoder—solid step toward learned interfaces. Yet it still relies on latent conditioning and leaves open the problems of retrieval control, editing without corruption, and long-horizon continuity.
PDM solves these at the architectural level:
9-node pressure signatures time for durable, importance-ranked storage
Resonance gating 3-vector triangulation for verifiable retrieval
Blip Proxy for minimal, interrupt-driven context injection during inference
No weight edits, no reconstruction loss, native export
The field is converging on external memory as the path to dependable agents. PDM is already the production implementation. Paper link in prior thread for contrast.
#PDM #MemoryArchitecture #ExternalMemory
likely latent state and hierarchy around the ar core*. i wouldn't try to ditch that at this point lol.
some other interesting bits which make me feel hmm:
like the Larimar paper (Larimar: BERT-style encoder writes facts into external memory, whose readout conditions GPT-2 or a GPT-style decoder without weight edits.)
and various steering/interp papers using or manipulating reps at various levels to augment the ar core and manipulate the residual stream
Larimar is another reminder that memory can be a learned memory interface around the AR core: write/update/forget mechanisms whose readouts condition generation.
===
>Larimar uses a BERT-style encoder during training and memory writing, but the decoder/base LM is not updated during fact editing. The “memory” gets written/updated, then its readout conditions the decoder. The Larimar paper had three modules: encoder, associative memory, decoder, trained together; then new facts can be added in one shot without retraining/fine-tuning the LLM.