# A Storage-Retrieval Gap in Parametric Knowledge Graph Memory

arXiv: https://arxiv.org/abs/2608.25489 · HTML: https://mlsanity.com/paper/2608.25489/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Knowledge-graph and structured-knowledge RAG, one of 15 directions in Retrieval-augmented generation.
- Builds on 6 of its 8 references that are in the map.
- First posted on arXiv 26 Aug 2026.

## Placed in

- Retrieval-augmented generation / [Knowledge-graph and structured-knowledge RAG](https://mlsanity.com/rag/knowledge-graph-and-structured-knowledge-rag/)

## Builds on (most cited first, 6 of 6)

- [QLoRA: Efficient Finetuning of Quantized LLMs](https://mlsanity.com/paper/2305.14314/) (2023)
- [Editing Models with Task Arithmetic](https://mlsanity.com/paper/2212.04089/) (2022)
- [Variational Reasoning for Question Answering with Knowledge Graph](https://mlsanity.com/paper/1709.04071/) (2017)
- [Parametric Retrieval Augmented Generation](https://mlsanity.com/paper/2501.15915/) (2025)
- [Dynamic Parametric Retrieval Augmented Generation for Test-time Knowledge Enhancement](https://mlsanity.com/paper/2503.23895/) (2025)
- [Parametric Retrieval-Augmented Generation using Latent Routing of LoRA Adapters](https://mlsanity.com/paper/2511.17044/) (2025)

## Abstract (arXiv)

Graph retrieval-augmented generation places retrieved subgraphs into the model's context window at query time, paying a recurring token cost and exposing source data on every call. We study an alternative: compiling a knowledge graph offline into a bank of LoRA adapters, one per entity, that serve as a parametric knowledge layer queried by injecting weights rather than text, at zero query-time context cost. On the MetaQA dataset, we find that subgraph-trained adapters encode context-free factual knowledge that generalizes to unseen questions: on single-valued relations the adapter gains +0.243 exact-match score over a base model that is nearly blind closed-book (0.007), and only the correct adapter recovers this knowledge (an oracle gap of +0.283 over the base model). However, the stored knowledge is not recoverable by similarity: given a query with no subgraph, embedding-based and weight-space geometry retrieval both perform at chance, because a semantically neighbouring entity's adapter does not contain the answer - knowledge is stored locally and does not transfer. Weight geometry correlates with subgraph semantics (ρ = +0.329) but not with functional retrievability. We quantify the byte and context-token costs against graph retrieval-augmented generation and discuss deployment implications. Our results establish that parametric knowledge graph memory is feasible for storing knowledge, and identify selecting and composing the right adapters by a mechanism other than semantic similarity as the central open problem - motivating a learned, query-conditioned composition mechanism.