新研究增强扩散语言模型的效率和语义 · 跟踪3个来源

报道来源 [3]

1. arXiv cs.AI TIER_1 English(EN) · Yeachan Jun, Albert No · 2026-07-21 04:00

   JUMP：对微调扩散语言模型的单次成员推理

   arXiv:2607.16207v1 Announce Type: new Abstract: Membership inference attacks (MIAs) test whether a candidate example appeared in a model's training data. We study MIAs for fine-tuned discrete diffusion language models (dLLMs), where membership means inclusion in the target model'…

2. arXiv cs.AI TIER_1 English(EN) · Ian Li, Zilei Shao, Benjie Wang, Rose Yu, Guy Van den Broeck, Anji Liu · 2026-07-21 04:00

   打破扩散语言模型中的因子分解壁垒

   arXiv:2603.00045v3 Announce Type: replace-cross Abstract: Diffusion language models theoretically allow for efficient parallel generation but are practically hindered by the ``factorization barrier'': the assumption that simultaneously predicted tokens are independent. This limit…

3. arXiv cs.CV TIER_1 English(EN) · Giorgos Petsangourakis, Christos Sgouropoulos, Bill Psomas, Theodoros Giannakopoulos, Giorgos Sfikas, Ioannis Kakogeorgiou · 2026-07-21 04:00

   REGLUE 您的潜在表示，利用全局和局部语义实现纠缠扩散

   arXiv:2512.16636v2 Announce Type: replace Abstract: Latent diffusion models (LDMs) achieve state-of-the-art image synthesis, yet their reconstruction-style denoising objective provides only indirect semantic supervision: high-level semantics emerge slowly, requiring longer traini…