LLM-Guided Co-Training for Text Classification

2025-09-19

AI Summary

To address the low utilization efficiency of unlabeled data in semi-supervised text classification, this paper proposes an LLM-guided weighted co-training method. The approach leverages large language models (LLMs) to generate high-quality pseudo-labels for unlabeled samples and introduces a dual-encoder architecture that dynamically estimates sample-level confidence scores to derive importance weights. These weights are exchanged bidirectionally between the two encoders and incorporated into backward propagation, enabling mutual knowledge distillation and training stability. The core innovation lies in treating the LLM as an interpretable knowledge amplifier, tightly coupling pseudo-label generation with adaptive weighting—thereby significantly enhancing the robustness and generalization of co-training under data-scarce conditions. Experiments on five benchmark datasets show state-of-the-art (SOTA) performance on four; Friedman’s test ranks the method first overall among 14 semi-supervised approaches, demonstrating statistically significant superiority over conventional methods.

Technology Category

• Machine Learning: Semi-Supervised Learning
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

In this paper, we introduce a novel weighted co-training approach that is guided by Large Language Models (LLMs). Namely, in our co-training approach, we use LLM labels on unlabeled data as target labels and co-train two encoder-only based networks that train each other over multiple iterations: first, all samples are forwarded through each network and historical estimates of each network's confidence in the LLM label are recorded; second, a dynamic importance weight is derived for each sample according to each network's belief in the quality of the LLM label for that sample; finally, the two networks exchange importance weights with each other -- each network back-propagates all samples weighted with the importance weights coming from its peer network and updates its own parameters. By strategically utilizing LLM-generated guidance, our approach significantly outperforms conventional SSL methods, particularly in settings with abundant unlabeled data. Empirical results show that it achieves state-of-the-art performance on 4 out of 5 benchmark datasets and ranks first among 14 compared methods according to the Friedman test. Our results highlight a new direction in semi-supervised learning -- where LLMs serve as knowledge amplifiers, enabling backbone co-training models to achieve state-of-the-art performance efficiently.

Problem

• Leveraging LLM labels for co-training text classification models
• Developing dynamic importance weighting between peer networks
• Enhancing semi-supervised learning with LLM-guided knowledge amplification

Innovation

• LLM-guided weighted co-training for classification
• Dynamic importance weights from peer network confidence
• LLMs as knowledge amplifiers in semi-supervised learning