A Dual-Axis Taxonomy of Knowledge Editing for LLMs: From Mechanisms to Functions

2025-08-12

AI Summary

To address the problem of knowledge obsolescence and inefficient updating in large language models (LLMs), this paper proposes the first taxonomy of knowledge editing that jointly considers *mechanisms* (e.g., parameter modification, external memory) and *functions* (e.g., factual, temporal, conceptual knowledge), formally defining edit objectives and corresponding evaluation tasks. Unlike prior taxonomies focusing solely on editing mechanisms, our framework systematically incorporates knowledge functionality as a core analytical dimension. Through a comprehensive survey of 120+ editing methods, we construct a unified framework spanning editing strategies, knowledge categories, and evaluation benchmarks. Our analysis reveals systematic trade-offs: distinct knowledge functions critically influence editing robustness, generalization, and side-effect profiles. The work clarifies the applicability boundaries of existing approaches, identifies *functional misalignment*—the mismatch between editing mechanisms and target knowledge functionality—as a fundamental bottleneck, and highlights key open challenges, including interpretable editing, cross-functional transfer, and dynamic knowledge evolution.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion
• Knowledge Representation and Reasoning: Knowledge Engineering

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Large language models (LLMs) acquire vast knowledge from large text corpora, but this information can become outdated or inaccurate. Since retraining is computationally expensive, knowledge editing offers an efficient alternative -- modifying internal knowledge without full retraining. These methods aim to update facts precisely while preserving the model's overall capabilities. While existing surveys focus on the mechanism of editing (e.g., parameter changes vs. external memory), they often overlook the function of the knowledge being edited. This survey introduces a novel, complementary function-based taxonomy to provide a more holistic view. We examine how different mechanisms apply to various knowledge types -- factual, temporal, conceptual, commonsense, and social -- highlighting how editing effectiveness depends on the nature of the target knowledge. By organizing our review along these two axes, we map the current landscape, outline the strengths and limitations of existing methods, define the problem formally, survey evaluation tasks and datasets, and conclude with open challenges and future directions.

Problem

• Updating outdated or inaccurate knowledge in LLMs efficiently
• Classifying knowledge editing methods by function and mechanism
• Evaluating editing effectiveness across diverse knowledge types

Innovation

• Dual-axis taxonomy for knowledge editing
• Function-based classification of knowledge types
• Mechanism-function mapping for editing effectiveness