Memory capacity is a persistent issue with large language models. They can struggle with long input sequences, thanks to the high cost of memory required by these models. Their training data can also become quickly obsolete as the world turns and…
Article Source
https://research.ibm.com/blog/memory-augmented-LLMs