MeMo: Memory as a Model

A new paradigm treats memory not as passive retrieval or LLM fine-tuning, but as a separate, fully trainable model that can be plugged into any existing large language model.

MeMo introduces a modular framework with a novel data synthesis pipeline, model merging for continual learning, and a black-box inference protocol. It delivers strong performance on complex benchmarks like BrowseComp-Plus, NarrativeQA, and MuSiQue while keeping inference costs constant and remaining robust to noisy data.

This shifts memory from a static lookup table into a dynamic, updatable component that evolves independently of the base model.

This Ledger Entry expands how we think about scalable knowledge in AI systems by showing how treating memory as its own trainable model solves the core trade-offs of RAG and continual fine-tuning.