Saved in:
Bibliographic Details
Main Authors: Tiwari, Sunil, Fofadiya, Payal
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.29194
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917372380577792
author Tiwari, Sunil
Fofadiya, Payal
author_facet Tiwari, Sunil
Fofadiya, Payal
contents Long-horizon dialogue systems suffer from semanticdrift and unstable memory retention across extended sessions. This paper presents a Multi-Layer Memory Framework that decomposes dialogue history into working, episodic, and semantic layers with adaptive retrieval gating and retention regularization. The architecture controls cross-session drift while maintaining bounded context growth and computational efficiency. Experiments on LOCOMO, LOCCO, and LoCoMo show improved performance, achieving 46.85 Success Rate, 0.618 overall F1 with 0.594 multi-hop F1, and 56.90% six-period retention while reducing false memory rate to 5.1% and context usage to 58.40%. Results confirm enhanced long-term retention and reasoning stability under constrained context budgets.
format Preprint
id arxiv_https___arxiv_org_abs_2603_29194
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Multi-Layered Memory Architectures for LLM Agents: An Experimental Evaluation of Long-Term Context Retention
Tiwari, Sunil
Fofadiya, Payal
Computer Vision and Pattern Recognition
Artificial Intelligence
Long-horizon dialogue systems suffer from semanticdrift and unstable memory retention across extended sessions. This paper presents a Multi-Layer Memory Framework that decomposes dialogue history into working, episodic, and semantic layers with adaptive retrieval gating and retention regularization. The architecture controls cross-session drift while maintaining bounded context growth and computational efficiency. Experiments on LOCOMO, LOCCO, and LoCoMo show improved performance, achieving 46.85 Success Rate, 0.618 overall F1 with 0.594 multi-hop F1, and 56.90% six-period retention while reducing false memory rate to 5.1% and context usage to 58.40%. Results confirm enhanced long-term retention and reasoning stability under constrained context budgets.
title Multi-Layered Memory Architectures for LLM Agents: An Experimental Evaluation of Long-Term Context Retention
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2603.29194