Larimar: Large Language Models with Episodic Memory Control
This addresses the challenge of knowledge updating in LLMs for AI practitioners, offering a more efficient alternative to re-training, though it is incremental as it builds on existing memory-augmented methods.
The paper tackles the problem of efficiently updating knowledge in Large Language Models (LLMs) by introducing Larimar, a brain-inspired architecture with distributed episodic memory that enables dynamic, one-shot updates without re-training. Experimental results show it achieves accuracy comparable to baselines while providing 8-10x speed-ups and flexibility for tasks like fact editing and forgetting.
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