面对被污染的观测数据，如何训练出干净的扩散模型？这篇论文给出了基于EM算法的全新方案，为生成模型在噪声场景下的应用打开了新思路。
arXiv:2407.01014v2 Announce Type: replace Abstract: Diffusion models excel in solving imaging inverse problems due to their ability to model complex i…
一篇论文提出了图概率模型来形式化LLM原生软件系统中的生成流，为开发者提供了可解释的定量分析框架。
arXiv:2606.15943v1 Announce Type: cross Abstract: Engineering LLM-native software remains a challenging and immature field. Current practice is largel…
将AI比作掷出无限可能性的骰子，引入时间维度计算来驾驭随机性，探索更智能的决策路径。
arXiv:2605.01134v2 Announce Type: replace Abstract: The dominant noun-based modeling paradigm has fundamentally constrained AI development, precluding…
新VAE变体有效处理重尾数据，Phase-Type分布带来创新突破
arXiv:2603.01800v2 Announce Type: replace-cross Abstract: Heavy-tailed distributions are ubiquitous in real-world data, where rare but extreme events …
提出概率递归意图切换模型，用概率方法建模意图动态转换，为序列决策和时序预测提供新思路
arXiv:2605.26998v1 Announce Type: new Abstract: Inverse reinforcement learning (IRL) recovers reward functions from observed behavior, yet traditional…
AI的本质一直是概率问题？这篇讨论从序列模型到通用AI的模拟路径，引发对统计学习深层的反思。
AI was always a probability problem.. If we look at the emergence of sequence models (or statistical learning in broader sense), they predict the next…
突破性研究：概率化小递归模型TRM，解决确定性递归陷入次优解难题，参数远少于大模型。
arXiv:2605.19943v1 Announce Type: new Abstract: Tiny Recursive Models (TRM) solve complex reasoning tasks with a fraction of the parameters of modern …
Njord首次实现海洋概率预报，结合图神经网络与潜变量框架，支持全球/区域场景。
arXiv:2605.15470v1 Announce Type: new Abstract: Ocean dynamics are inherently chaotic, yet existing machine learning ocean models produce only determi…