LC-ICL: Label-Guided Contrastive In-Context Learning for Robust Information Extraction

2026-06-28

AI Summary

Current large language models for few-shot information extraction rely solely on positive examples for in-context learning, overlooking the discriminative signals embedded in negative examples, which compromises model robustness. This work proposes a label-guided contrastive in-context learning approach that, for the first time, incorporates negative examples annotated with error-reason labels. By constructing demonstration examples that pair hard negatives with nearest-neighbor positives, the method enhances the model’s ability to recognize and avoid common error patterns. Evaluated on named entity recognition and relation extraction tasks, the proposed approach significantly outperforms existing few-shot in-context learning methods, demonstrating superior generalization and adaptability across multiple datasets.

Abstract

There has been increasing interest in exploring the capabilities of advanced large language models (LLMs) in the field of information extraction (IE), specifically focusing on tasks related to named entity recognition (NER) and relation extraction (RE).Although researchers are exploring the use of few-shot information extraction through in-context learning with LLMs, they tend to focus only on using correct or positive examples for demonstration, neglecting the potential value of incorporating incorrect or negative examples into the learning process.In this paper, we present LC-ICL a novel few-shot technique that leverages both correct and incorrect sample constructions to create in-context learning demonstrations. This approach enhances the ability of LLMs to extract entities and relations by combining positive samples with negative samples annotated by error-cause labels. These labels expose more detailed error features in erroneous examples, enabling the model to understand why similar predictions fail and avoid repeating such errors during inference.Specifically, our proposed method taps into the inherent contextual information and valuable information in hard negative samples and the nearest positive neighbors to the test and then applies the in-context learning demonstrations based on LLMs. Our experiments on various datasets indicate that LC-ICL outperforms previous few-shot in-context learning methods, delivering substantial enhancements in performance across a broad spectrum of related tasks. These improvements are noteworthy, showcasing the versatility of our approach in diverse scenarios.

Problem

• information extraction
• in-context learning
• few-shot learning
• negative examples
• large language models

Innovation

• in-context learning
• contrastive learning
• label-guided
• few-shot information extraction
• error-cause labeling