Abstract:Large language models (LLMs) achieve strong performance across a wide range of tasks, but remain frozen after pretraining until subsequent updates. Many real-world applications require timely, domain-specific information, motivating the need for efficient mechanisms to incorporate new knowledge. In this paper, we introduce MeMo (Memory as a Model), a modular framework that encodes new knowledge into a dedicated memory model while keeping the LLM parameters unchanged. Compared to existing methods, MeMo offers several advantages: (a) it captures complex cross-document relationships, (b) it is robust to retrieval noise, (c) it avoids catastrophic forgetting in the LLM, (d) it does not require access to the LLM's weights or output logits, enabling plug-and-play integration with both open and proprietary closed-source LLMs, and (e) its retrieval cost is independent of corpus size at inference time. Our experimental results on three benchmarks, BrowseComp-Plus, NarrativeQA, and MuSiQue, show that MeMo achieves strong performance compared to existing methods across diverse settings.
| Comments: | This paper introduces MeMo, a framework that augments any LLM with up-to-date or domain-specific knowledge via a trained memory model, avoiding costly retraining, mitigating catastrophic forgetting, and remaining robust to retrieval noise |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2605.15156 [cs.CL] |
| (or arXiv:2605.15156v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2605.15156 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Arun Verma [view email]
[v1]
Thu, 14 May 2026 17:51:34 UTC (490 KB)