Human-like Episodic Memory for Infinite Context LLMs
Zafeirios Fountas, Martin A Benfeghoul, Adnan Oomerjee, Fenia Christopoulou, Gerasimos Lampouras, Haitham Bou-Ammar, Jun Wang
2024-07-15
Summary
This paper presents EM-LLM, a new approach that enhances large language models (LLMs) by integrating a human-like episodic memory system, allowing them to handle very long contexts more effectively.
What's the problem?
While LLMs are powerful, they struggle to maintain coherence and accuracy when dealing with long sequences of text. This limitation makes it difficult for them to remember and relate information over extended conversations or documents, which is something humans do naturally with their episodic memory.
What's the solution?
EM-LLM addresses this issue by mimicking how humans organize and retrieve memories. It organizes sequences of information into coherent events and uses a two-stage memory process to retrieve these events based on similarity and time. This allows the model to efficiently access relevant information, even from very long texts. The researchers tested EM-LLM on the LongBench dataset and found that it outperformed existing models in various tasks, showing significant improvements in performance.
Why it matters?
This research is important because it not only improves the capabilities of LLMs in processing long contexts but also provides insights into how human memory works. By bridging the gap between artificial intelligence and human cognitive processes, EM-LLM opens up new possibilities for AI applications in areas like education, customer service, and any field where understanding complex information over time is crucial.
Abstract
Large language models (LLMs) have shown remarkable capabilities, but still struggle with processing extensive contexts, limiting their ability to maintain coherence and accuracy over long sequences. In contrast, the human brain excels at organising and retrieving episodic experiences across vast temporal scales, spanning a lifetime. In this work, we introduce EM-LLM, a novel approach that integrates key aspects of human episodic memory and event cognition into LLMs, enabling them to effectively handle practically infinite context lengths while maintaining computational efficiency. EM-LLM organises sequences of tokens into coherent episodic events using a combination of Bayesian surprise and graph-theoretic boundary refinement in an on-line fashion. When needed, these events are retrieved through a two-stage memory process, combining similarity-based and temporally contiguous retrieval for efficient and human-like access to relevant information. Experiments on the LongBench dataset demonstrate EM-LLM's superior performance, outperforming the state-of-the-art InfLLM model with an overall relative improvement of 4.3% across various tasks, including a 33% improvement on the PassageRetrieval task. Furthermore, our analysis reveals strong correlations between EM-LLM's event segmentation and human-perceived events, suggesting a bridge between this artificial system and its biological counterpart. This work not only advances LLM capabilities in processing extended contexts but also provides a computational framework for exploring human memory mechanisms, opening new avenues for interdisciplinary research in AI and cognitive science.