Revealing and Mitigating Over-Attention in Knowledge Editing
This addresses a critical issue for researchers and practitioners using knowledge editing to correct errors in LLMs, though it is incremental as it builds on existing editing methods.
The paper tackles the problem of Specificity Failure in knowledge editing for Large Language Models, where editing specific knowledge inadvertently corrupts other pre-existing knowledge, and introduces the Selective Attention Drift Restriction (SADR) method to mitigate this by restricting attention weight changes, showing significant mitigation in experiments on five LLMs.
Large Language Models have demonstrated superior performance across a wide range of tasks, but they still exhibit undesirable errors due to incorrect knowledge learned from the training data. To avoid this, knowledge editing methods emerged to precisely edit the specific model knowledge via efficiently modifying a very small percentage of parameters. % However, those methods can lead to the problem of Specificity Failure: when the content related to the edited knowledge occurs in the context, it can inadvertently corrupt other pre-existing knowledge. However, those methods can lead to the problem of Specificity Failure, where the existing knowledge and capabilities are severely degraded due to editing. Our preliminary indicates that Specificity Failure primarily stems from the model's attention heads assigning excessive attention scores to entities related to the edited knowledge, thereby unduly focusing on specific snippets within the context, which we denote as the Attention Drift phenomenon. To mitigate such Attention Drift issue, we introduce a simple yet effective method Selective Attention Drift Restriction}(SADR), which introduces an additional regularization term during the knowledge editing process to restrict changes in the attention weight distribution, thereby preventing undue focus on the edited entity. Experiments on five frequently used strong LLMs demonstrate the effectiveness of our method, where SADR can significantly mitigate Specificity Failure in the predominant knowledge editing tasks.