EUVIMEDEUVIMED entdecken
Crossref

When AUC Meets Generative Reconstruction: A Unified Principle for Balanced Visual Representation Learning

Boyu Han, Qianqian Xu, Shilong Bao, Zhiyong Yang, Qingming Huang

IEEE Transactions on Pattern Analysis and Machine Intelligence · 2026 · 1-18

DOI: 10.1109/tpami.2026.3754040 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026. Verfasst von Boyu Han und 4 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Multimodal large language models (MLLMs) require visual encoders that extract fine-grained visual features while constructing semantically disentangled and discriminative representations. Conventional methodologies encounter a fundamental trade-off: contrastive learning enhances inter-class discrimination by suppressing intra-class variability, whereas generative reconstruction emphasizes perceptual fidelity but lacks explicit comparative supervision. The naive aggregation of these objectives via composite loss functions often induces optimization conflicts and instability due to gradient interference within the shared latent space. To resolve this, we introduce a unified representation-learning paradigm that reformulates generative reconstruction as a discriminative ranking problem within the generative output manifold. Our approach maintains perceptual reconstruction as a supervisory signal and integrates a ranking-based discriminative constraint by systematically ordering matched and mismatched reconstruction outputs. This formulation embeds discriminative information seamlessly into the generative process, alleviating the risk of loss competition. Specifically, we propose AUC Generative Reconstruction (AUCGR), which optimizes the area under the ROC curve (AUC) by ensuring higher scores for matched pairs than for mismatched pairs, and Partial AUC Generative Reconstruction (pAUCGR), which addresses supervision bias by focusing on mismatched comparisons within reliable feature subspaces. The proposed objectives are designed to be plug-and-play, requiring no architectural modifications to existing generative reconstruction frameworks. Theoretical analysis establishes rigorous connections to AUC/pAUC optimization and provides generalization guarantees. Extensive experiments across diverse CLIP architectures, more than 30 visual benchmarks, generative backbones, and MLLMs consistently validate that our method advances both discriminative accuracy and fine-grained perceptual representation, establishing a new benchmark for visual representation learning.

Zitieren

APA

Han, B., Xu, Q., Bao, S., Yang, Z., & Huang, Q. (2026). When AUC Meets Generative Reconstruction: A Unified Principle for Balanced Visual Representation Learning. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1-18. https://doi.org/10.1109/tpami.2026.3754040

Vancouver

Han B, Xu Q, Bao S, Yang Z, Huang Q. When AUC Meets Generative Reconstruction: A Unified Principle for Balanced Visual Representation Learning. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2026:1-18. doi:10.1109/tpami.2026.3754040

BibTeX
@article{Han2026,
  author = {Han, Boyu and Xu, Qianqian and Bao, Shilong and Yang, Zhiyong and Huang, Qingming},
  title = {When AUC Meets Generative Reconstruction: A Unified Principle for Balanced Visual Representation Learning},
  journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year = {2026},
  pages = {1--18},
  doi = {10.1109/tpami.2026.3754040}
}
Kostenlos starten →