Larimar: Large Language Models with Episodic Memory Control

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Das, Payel, Chaudhury, Subhajit, Nelson, Elliot, Melnyk, Igor, Swaminathan, Sarath, Dai, Sihui, Lozano, Aurélie, Kollias, Georgios, Chenthamarakshan, Vijil, Jiří, Navrátil, Dan, Soham, Chen, Pin-Yu
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914920064352256
author Das, Payel
Chaudhury, Subhajit
Nelson, Elliot
Melnyk, Igor
Swaminathan, Sarath
Dai, Sihui
Lozano, Aurélie
Kollias, Georgios
Chenthamarakshan, Vijil
Jiří
Navrátil
Dan, Soham
Chen, Pin-Yu
author_facet Das, Payel
Chaudhury, Subhajit
Nelson, Elliot
Melnyk, Igor
Swaminathan, Sarath
Dai, Sihui
Lozano, Aurélie
Kollias, Georgios
Chenthamarakshan, Vijil
Jiří
Navrátil
Dan, Soham
Chen, Pin-Yu
contents Efficient and accurate updating of knowledge stored in Large Language Models (LLMs) is one of the most pressing research challenges today. This paper presents Larimar - a novel, brain-inspired architecture for enhancing LLMs with a distributed episodic memory. Larimar's memory allows for dynamic, one-shot updates of knowledge without the need for computationally expensive re-training or fine-tuning. Experimental results on multiple fact editing benchmarks demonstrate that Larimar attains accuracy comparable to most competitive baselines, even in the challenging sequential editing setup, but also excels in speed - yielding speed-ups of 8-10x depending on the base LLM - as well as flexibility due to the proposed architecture being simple, LLM-agnostic, and hence general. We further provide mechanisms for selective fact forgetting, information leakage prevention, and input context length generalization with Larimar and show their effectiveness. Our code is available at https://github.com/IBM/larimar
format Preprint
id arxiv_https___arxiv_org_abs_2403_11901
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Larimar: Large Language Models with Episodic Memory Control
Das, Payel
Chaudhury, Subhajit
Nelson, Elliot
Melnyk, Igor
Swaminathan, Sarath
Dai, Sihui
Lozano, Aurélie
Kollias, Georgios
Chenthamarakshan, Vijil
Jiří
Navrátil
Dan, Soham
Chen, Pin-Yu
Machine Learning
Artificial Intelligence
Efficient and accurate updating of knowledge stored in Large Language Models (LLMs) is one of the most pressing research challenges today. This paper presents Larimar - a novel, brain-inspired architecture for enhancing LLMs with a distributed episodic memory. Larimar's memory allows for dynamic, one-shot updates of knowledge without the need for computationally expensive re-training or fine-tuning. Experimental results on multiple fact editing benchmarks demonstrate that Larimar attains accuracy comparable to most competitive baselines, even in the challenging sequential editing setup, but also excels in speed - yielding speed-ups of 8-10x depending on the base LLM - as well as flexibility due to the proposed architecture being simple, LLM-agnostic, and hence general. We further provide mechanisms for selective fact forgetting, information leakage prevention, and input context length generalization with Larimar and show their effectiveness. Our code is available at https://github.com/IBM/larimar
title Larimar: Large Language Models with Episodic Memory Control
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2403.11901