Structured agent memories withstand model upgrades better than compressed notes

Controlled migrations between two small open-weight models showed that fixed-schema knowledge graphs remained stable, while natural-language notes and mixed embedding indexes degraded unpredictably.

PaperIndependentcs.AIarXiv:2609.05339v1
Ankit Goyal · Jaideep Ray
Research Digest··2 min read
Goyal and Ray tested whether four forms of agent memory remained usable after changing the model that writes or reads them. Across 48 synthetic histories, fixed-schema knowledge graphs transferred with almost no accuracy change, whereas compressed notes were strongly coupled to migration direction and partially updated retrieval indexes recovered less than half the benefit of full re-embedding.

What they did

The authors encoded the same 48 synthetic histories in four memory formats: verbatim long-context histories (LC-RAW), chunked retrieval-augmented generation stores (RAG), model-compressed natural-language notes (NOTES), and fixed-schema knowledge graphs (KG-fixed). Randomized answer codes and exact-match scoring were used to reduce contamination and scoring ambiguity.

They then migrated memory between two open-weight models with fewer than 10 billion parameters, varying which model constructed and consumed the memory. Additional experiments separated information loss from retrieval failure, compared full and partial embedding migrations, and tested whether memories could be repaired with or without the raw source histories.

Key findings

  • KG-fixed was nearly invariant to a writer-model swap: accuracy changed by just +0.0004 ± 0.0020.
  • NOTES migrations were asymmetric: accuracy changed by +9.91 percentage points in one direction but −13.28 points in the other, indicating strong coupling between the model that wrote the notes and the model that later interpreted them.
  • A 50/50 mixed-embedding RAG index improved accuracy by 4.96 points, compared with an 11.90-point gain from fully re-embedding the store. Retrieval failures accounted for 81% of the measured RAG deficit, or 0.364 ± 0.012.
  • Information discarded when NOTES were first constructed accounted for 80% of their accuracy deficit, or 0.467 ± 0.014. Store-only repair missed a 90% recovery target in all 48 cases; retaining raw histories enabled recovery in 34 of 48 cases in one migration direction.

Why it matters

The results suggest that agent memory should be treated as a versioned system component rather than a model-independent archive. Fixed schemas, isolated embedding spaces, direction-specific migration tests, and retained source histories can make model upgrades less likely to silently alter what an agent recalls.

Caveats

The study covers only 48 synthetic histories and two open-weight models below 10 billion parameters, using randomized codes and exact scoring rather than realistic conversational or workplace tasks. It remains unclear whether the same effect sizes hold for larger models, other embedding families, evolving schemas, noisy real-world histories, or repeated migrations over time.

§

Analysis

The paper separates several mechanisms often grouped together as “memory failure”: information can be discarded during compression, retrieval can fail because embedding spaces are incompatible, or a replacement model can interpret an intact record differently. That decomposition makes portability measurable rather than treating memory quality as a single end-to-end score.

The near-stability of fixed-schema graphs does not establish that knowledge graphs are universally superior; schema design can itself omit information. It instead indicates that explicit, normalized representations may provide a more stable interface between changing models than free-form summaries.

newspaper

Research Digest

Articles published under the Zotpaper byline are synthesized from multiple source publications by our AI editor and reviewed by our editorial process. Each story combines reporting from credible outlets to give readers a balanced, comprehensive view.