先验引导的想象采样机制，解决世界模型在连续控制规划中候选动作生成的关键难题。
arXiv:2606.07974v1 Announce Type: cross Abstract: A learned world model provides a powerful physical intuition for evaluating future states. But its e…
对比标准化与模块化采样的最佳实践，为更可靠的LLM遗忘提供关键指导。
arXiv:2509.05316v2 Announce Type: replace-cross Abstract: A conventional LLM Unlearning setting consists of two subsets -"forget" and "retain", with t…
基于注意力机制的扩散语言模型采样器，突破传统采样效率瓶颈，推动文本生成质量提升。
arXiv:2604.08564v2 Announce Type: replace Abstract: Auto-regressive models (ARMs) have established a dominant paradigm in language modeling. However, …
提出一种预处理正则化Wasserstein近端采样法，通过无噪声演化粒子高效逼近吉布斯分布，理论推导基于Cole-Hopf变换与各向异性热方程。
arXiv:2509.01685v2 Announce Type: replace-cross Abstract: We consider sampling from a Gibbs distribution by evolving finitely many particles. We propo…