LoFT: Parameter-Efficient Fine-Tuning for Long-tailed Semi-Supervised Learning in Open-World Scenarios

2025-09-11

AI Summary

To address challenges in open-world long-tailed semi-supervised learning (LTSSL)—including low-quality pseudo-labels, model overconfidence, and interference from out-of-distribution (OOD) samples—this paper proposes LoFT, a parameter-efficient fine-tuning framework, and its extension LoFT-OW. LoFT pioneers the integration of foundation models into the LTSSL paradigm, employing lightweight adaptation to generate high-confidence, high-accuracy pseudo-labels. LoFT-OW further incorporates OOD detection and adaptive pseudo-label refinement, significantly enhancing robustness to unknown classes. The method achieves stable training using only 1% unlabeled data and consistently surpasses state-of-the-art methods across multiple long-tailed semi-supervised benchmarks. Experimental results demonstrate superior efficacy and generalization under both class imbalance and open-world conditions, validating the framework’s efficiency and adaptability in realistic, distribution-shifted scenarios.

Technology Category

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

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Long-tailed learning has garnered increasing attention due to its wide applicability in real-world scenarios. Among existing approaches, Long-Tailed Semi-Supervised Learning (LTSSL) has emerged as an effective solution by incorporating a large amount of unlabeled data into the imbalanced labeled dataset. However, most prior LTSSL methods are designed to train models from scratch, which often leads to issues such as overconfidence and low-quality pseudo-labels. To address these challenges, we extend LTSSL into the foundation model fine-tuning paradigm and propose a novel framework: LoFT (Long-tailed semi-supervised learning via parameter-efficient Fine-Tuning). We demonstrate that fine-tuned foundation models can generate more reliable pseudolabels, thereby benefiting imbalanced learning. Furthermore, we explore a more practical setting by investigating semi-supervised learning under open-world conditions, where the unlabeled data may include out-of-distribution (OOD) samples. To handle this problem, we propose LoFT-OW (LoFT under Open-World scenarios) to improve the discriminative ability. Experimental results on multiple benchmarks demonstrate that our method achieves superior performance compared to previous approaches, even when utilizing only 1% of the unlabeled data compared with previous works.

Problem

• Addresses overconfidence and low-quality pseudo-labels in long-tailed semi-supervised learning
• Extends LTSSL to foundation model fine-tuning paradigm for improved reliability
• Handles open-world scenarios with out-of-distribution samples in unlabeled data

Innovation

• Parameter-efficient fine-tuning for long-tailed learning
• Generating reliable pseudo-labels using foundation models
• Open-world adaptation handling out-of-distribution samples