从语义词典中学习：面向统一视觉表征与生成的判别式码本对比学习
Learning from Semantic Dictionaries: Discriminative Codebook Contrastive Learning for Unified Visual Representation and Generation
把生成式掩码词元重建与码本对比损失结合在同一个离散词元空间里训练，为视觉研究者提供了一条无需数据增强、教师模型和在线分词器即可同时获得强表征与高质量生成的统一预训练路径，可直接复用于统一理解-生成模型的骨干预训练。
原文依据 a masked token reconstruction loss that captures fine-grained generative detail, and a codebook contrast loss that aligns encoder features with discriminative semantics
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论判别式与生成式视觉模型在各自领域表现优异，但在语义上仍然错位，阻碍了统一视觉学习的进展。我们提出 LEASE（LEArning from SEmantic Dictionaries，从语义词典中学习），这是一个自监督框架，通过成对的生成-判别码本设计弥合这一鸿沟。LEASE 完全在通过一次性预计算步骤得到的离散词元空间中运行，从而无需数据增强、教师模型或在线分词器即可高效训练。LEASE 集成了两个互补的目标：一个掩码词元重建损失，用于捕捉细粒度的生成细节；以及一个码本对比损失，通过自适应质心加权使编码器特征与判别式语义对齐。这种双重监督产生了一个统一的潜空间，同时支持高质量生成与强表征学习。在 ImageNet-1K 上，LEASE 取得了最优的统一性能，在线性探测、无条件生成、少样本学习、迁移与鲁棒性基准上优于 MAGE 和 Sorcen 等先前基于 VQGAN 的方法。它也与领域专用的对比式与生成式模型相比具有竞争力，同时超越了先前的 MIM 方法。无监督的 LEASE 模型还可以在其学习到的表征基础上扩展至条件生成，并证明与专用基线相比具有竞争力。总体而言，LEASE 为联合理解与生成视觉内容的通用视觉模型迈出了高效而有效的一步。代码将在论文录用后发布。
英文原文摘要 1618 字符
Discriminative and generative vision models excel in their respective domains but remain semantically misaligned, hindering progress toward unified visual learning. We introduce LEASE (LEArning from SEmantic Dictionaries), a self-supervised framework that bridges this gap using a paired generative-discriminative codebook design. LEASE operates entirely in a discrete token space produced through a one-time precomputation step, enabling efficient training without data augmentations, teacher models, or online tokenizers. LEASE integrates two complementary objectives: a masked token reconstruction loss that captures fine-grained generative detail, and a codebook contrast loss that aligns encoder features with discriminative semantics via adaptive centroid weighting. This dual supervision yields a unified latent space that supports both high-quality generation and strong representation learning.On ImageNet-1K, LEASE achieves state-of-the-art unified performance, outperforming prior VQGAN-based methods such as MAGE and Sorcen across linear probing, unconditional generation, few-shot learning, transfer, and robustness benchmarks. It also competes favorably with domain-specialized contrastive and generative models while surpassing previous MIM methods. The unsupervised LEASE model can also be extended to conditional generation by building upon its learned representations, proving competitive with specialized baselines.Overall, LEASE provides an efficient and effective step toward general-purpose vision models that jointly understand and generate visual content. Code will be released upon acceptance.