In-Context Editing: Learning Knowledge from Self-Induced Distributions

2024-06-17

arXiv.org

AI Summary

To address overfitting, poor generalization, and unnatural generation in knowledge editing for large language models (LLMs), this paper proposes Consistent Contextual Editing (ICE). ICE reformulates knowledge editing as a contextual output distribution alignment task—departing from conventional pointwise label correction. It achieves fine-tuning-free, overfitting-resistant knowledge injection via gradient-driven distribution matching and self-induced contextual distribution modeling, without model retraining. The method inherently supports continual editing while preserving accuracy, locality, generalization, and linguistic fluency. Evaluated on four standard knowledge editing benchmarks, ICE consistently outperforms state-of-the-art baselines, significantly improving editing robustness and generation naturalness. Experimental results validate ICE as a new paradigm for efficient, trustworthy, and sustainable LLM knowledge updating.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Data Mining & Knowledge Management: Knowledge Acquisition from the Web

Application Category

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

Abstract

In scenarios where language models must incorporate new information efficiently without extensive retraining, traditional fine-tuning methods are prone to overfitting, degraded generalization, and unnatural language generation. To address these limitations, we introduce Consistent In-Context Editing (ICE), a novel approach leveraging the model's in-context learning capability to optimize toward a contextual distribution rather than a one-hot target. ICE introduces a simple yet effective optimization framework for the model to internalize new knowledge by aligning its output distributions with and without additional context. This method enhances the robustness and effectiveness of gradient-based tuning methods, preventing overfitting and preserving the model's integrity. We analyze ICE across four critical aspects of knowledge editing: accuracy, locality, generalization, and linguistic quality, demonstrating its advantages. Experimental results confirm the effectiveness of ICE and demonstrate its potential for continual editing, ensuring that the integrity of the model is preserved while updating information.

Problem

• Language Model
• Overlearning
• Text Generation

Innovation

• ICE method
• Fast Learning
• Overfitting Prevention