EvoEdit: Lifelong Free-Text Knowledge Editing through Latent Perturbation Augmentation and Knowledge-driven Parameter Fusion

2025-12-04

AI Summary

To address post-deployment knowledge obsolescence in large language models (LLMs), existing methods—relying on structured triple-based representations and one-shot editing paradigms—fail to support natural-language-based, continual knowledge updates. This paper introduces Lifetime Free-text Knowledge Editing (LF-Edit), a novel task enabling multi-turn, sustainable knowledge injection and forgetting mitigation directly from free-text inputs. To rigorously evaluate LF-Edit, we construct MRLF-Bench, a large-scale, multi-level benchmark, and propose a cognition-inspired four-tier evaluation framework: Memory, Comprehension, Constrained Comprehension, and Reasoning. We further develop Latent Perturbation Augmentation to improve injection fidelity and introduce a knowledge-driven parameter fusion mechanism to preserve historical knowledge stability. On MRLF-Bench, our approach significantly outperforms existing baselines, achieving a superior balance between acquiring new knowledge and retaining prior knowledge.

Technology Category

• Machine Learning: Life-Long and Continual Learning
• Natural Language Processing: (Large) Language Models
• Knowledge Representation and Reasoning: Knowledge Acquisition

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: Large language models for search

Abstract

Adjusting the outdated knowledge of large language models (LLMs) after deployment remains a major challenge. This difficulty has spurred the development of knowledge editing, which seeks to accurately and efficiently modify a model's internal (parametric) knowledge without retraining it from scratch. However, existing methods suffer from two limitations. First, they depend on structured triplets that are misaligned with the free-text nature of LLM pretraining and fail to capture the nuanced relationships among facts. Second, they typically support one-time knowledge updates, with relatively limited research on the problem of sequential or lifelong editing. To address these gaps, we propose a new task, Lifelong Free-text Knowledge Editing (LF-Edit), which enables models to incorporate updates expressed in natural language and supports continual editing over time. Despite its promise, LF-Edit faces the dual challenge of integrating new knowledge while mitigating the forgetting of prior information. To foster research on this new task, we construct a large-scale benchmark, Multi-Rank Lifelong Free-text Editing Benchmark (MRLF-Bench), containing 16,835 free-text edit requests. We further design a cognitively inspired multi-rank evaluation framework encompassing four levels: memorization, understanding, constrained comprehension, and reasoning. To tackle the challenges inherent in LF-Edit, we introduce a novel approach named EvoEdit that enhances knowledge injection through Latent Perturbation Augmentation and preserves prior information via Knowledge-driven Parameter Fusion. Experimental results demonstrate that EvoEdit substantially outperforms existing knowledge editing methods on the proposed LF-Edit task.

Problem

• Adjust outdated knowledge in large language models post-deployment
• Enable lifelong editing using natural language updates
• Mitigate forgetting of prior information during sequential editing

Innovation

• Latent Perturbation Augmentation for free-text knowledge injection
• Knowledge-driven Parameter Fusion to prevent forgetting prior information
• Multi-rank evaluation framework for lifelong editing benchmark