Latent Knowledge Scalpel: Precise and Massive Knowledge Editing for Large Language Models
This addresses the challenge of efficiently updating factual knowledge in LLMs for users needing reliable and up-to-date AI systems, though it builds incrementally on existing model editing methods.
The paper tackles the problem of inaccurate or outdated information in Large Language Models by introducing the Latent Knowledge Scalpel (LKS), which enables precise editing of up to 10,000 facts simultaneously while preserving the models' general capabilities, as demonstrated on Llama-2 and Mistral.
Large Language Models (LLMs) often retain inaccurate or outdated information from pre-training, leading to incorrect predictions or biased outputs during inference. While existing model editing methods can address this challenge, they struggle with editing large amounts of factual information simultaneously and may compromise the general capabilities of the models. In this paper, our empirical study demonstrates that it is feasible to edit the internal representations of LLMs and replace the entities in a manner similar to editing natural language inputs. Based on this insight, we introduce the Latent Knowledge Scalpel (LKS), an LLM editor that manipulates the latent knowledge of specific entities via a lightweight hypernetwork to enable precise and large-scale editing. Experiments conducted on Llama-2 and Mistral show even with the number of simultaneous edits reaching 10,000, LKS effectively performs knowledge editing while preserving the general abilities of the edited LLMs. Code is available at: https://github.com/Linuxin-xxx/LKS.