SECOS: Semantic Capture for Rigorous Classification in Open-World Semi-Supervised Learning

2026-04-30

AI Summary

This work addresses the challenge in open-world semi-supervised learning where existing methods struggle to perform precise semantic classification on unlabeled data containing both known and novel classes due to the absence of explicit semantic supervision and alignment mechanisms. To overcome this limitation, the authors propose SECOS, a novel approach that, for the first time in this setting, enables end-to-end prediction of textual labels without requiring post-processing. SECOS integrates external knowledge bases, aligns image and text semantics through cross-modal representation learning, and establishes a unified framework that jointly optimizes representation learning and semi-supervised training, thereby providing explicit semantic supervision for novel classes. Experimental results demonstrate that SECOS achieves up to a 5.4% performance gain over state-of-the-art methods even under more relaxed evaluation protocols, significantly advancing the frontier of open-world semantic classification.

Abstract

In open-world semi-supervised learning (OWSSL), a model learns from labeled data and unlabeled data containing both known and novel classes. In practical OWSSL applications, models are expected to perform rigorous classification by directly selecting the most semantically relevant label from a candidate set for each sample. Existing OWSSL methods fail to achieve this because novel samples are trained without explicit supervision, and these methods lack mechanisms to extract latent semantic information, resulting in predicted labels that have no semantic correspondence to candidate textual labels. To address this, we introduce SEmantic Capture for Open-world Semi-supervised learning (SECOS), which directly predicts textual labels from the candidate set without post-processing, meeting the requirements of practical OWSSL applications. SECOS leverages external knowledge to extract and align semantic representations across modalities for both known and novel classes, providing explicit supervisory signals for training novel classes. Extensive experiments demonstrate that even when existing OWSSL methods are evaluated under the more lenient post-hoc matching setting, SECOS still surpasses them by up to 5.4\% without such assistance, highlighting its superior effectiveness. Code is available at https://github.com/ganchi-huanggua/OSSL-Classification.

Problem

• open-world semi-supervised learning
• semantic classification
• novel class discovery
• textual label alignment
• rigorous classification

Innovation

• Semantic Capture
• Open-World Semi-Supervised Learning
• Cross-Modal Alignment
• Textual Label Prediction
• Explicit Supervision for Novel Classes