# Memorilla preprint reports improved retrieval using 16 memory tokens

_Published Friday, October 16, 2026 at 4:21 AM EDT · AI, Science · Latest · Tier 2 — Notable_

Researchers report that Memorilla, a memory system for language models with fixed parameters, outperformed baselines including Mem0 and MemoRAG on five of eight benchmarks using 16 memory tokens. The arXiv preprint covers personalization, long-document questions and knowledge-intensive questions.

Memorilla encodes documents once into reusable numerical representations, then learns to place query-specific information into the model's input. On FactKG, it reached 84.33% accuracy versus 80.43% with full-context inference, using 108x fewer tokens to represent source documents. In TextWorld, average agent success rose from 18.6% with full-history input to 24.2%; these results concern the evaluated benchmarks and environment.

## Sources

- [cs.LG updates on arXiv.org](https://arxiv.org/abs/2610.21375)

---
Canonical: https://techandbusiness.org/newswire/kZQHbh8nzH_WB6S-bZKF49
Published: 2026-10-16T08:21:31.510Z
Story chronology: 2026-10-16T04:00:00.000Z
Retrieved: 2026-10-16T11:31:54.982Z
Publisher: Tech & Business (techandbusiness.org)