AI TrackerFrontier desk · AEST
Back to news
normalaitech

EngramEdit: Decoupled Knowledge Updates in LLMs through Conditional Memory

arXiv cs.CL · 8 Oct 2026, 03:58 AEST

EngramEdit introduces a method for decoupled knowledge updates in large language models using conditional memory architectures. This approach allows for independent factual knowledge updates while preserving unrelated knowledge and general capabilities.

Key facts

Read aloud

EngramEdit introduces a method for decoupled knowledge updates in large language models using conditional memory architectures. This approach allows for independent factual knowledge updates while preserving unrelated knowledge and general capabilities. Conditional memory architectures such as DeepSeek Engram use input n-grams to look up learned embeddings, expanding the capacity of large language models L L M s with limited additional computation. EngramEdit first computes target memory representations that make the model predict the updated fact across multiple expressions. Experiments show that EngramEdit enables independent factual knowledge updates through conditional memory, achieving near-perfect editing success. Revised knowledge is usable across unseen expressions and in multi-hop reasoning, with nearly three times the strongest baseline's accuracy under chain-of-thought CoT prompting.

Sources in this cluster