最大开源阿拉伯语大模型家族Jais 2发布，专精阿拉伯文化场景，性能对标国际主流模型。
arXiv:2608.13580v1 Announce Type: cross Abstract: Jais 2 is a family of Arabic-centric large language models developed jointly by MBZUAI, Cerebras, an…
利用Stein散度提升安全强化学习对尾部风险的敏感度，UAI 2026收录新方法
arXiv:2607.13175v1 Announce Type: cross Abstract: Safe reinforcement learning typically enforces safety by bounding expected cumulative costs, a crite…
共形预测遇上认知不确定性，新方法在预测区间上实现最优校准，UAI 2026收录的前沿研究。
arXiv:2505.19033v2 Announce Type: replace-cross Abstract: Conformal prediction (CP) is a widely used frequentist framework to quantify uncertainty by …
揭示大语言模型在线自我改进对齐的收敛性理论，为LLM安全与性能优化提供数学基础，UAI 2026收录。
arXiv:2606.31524v1 Announce Type: cross Abstract: The Self-Improving Alignment (SAIL) algorithm addresses distribution shift by reducing a bilevel for…
因果图未知时如何保证公平？这项研究在变量簇上实现了干预公平，为AI可靠性提供新解法。
arXiv:2602.23611v2 Announce Type: replace-cross Abstract: Algorithmic decisions about individuals require predictions that are not only accurate but a…
颠覆传统AI范式，提出不使用模型即可实现通用智能的激进新理论，来自UAI 2026的最新研究
arXiv:2602.23242v3 Announce Type: replace Abstract: In general reinforcement learning, all established optimal agents, including AIXI, are model-based…
利用软提示蒸馏技术，在设备端实现安全大模型系统，兼顾效率与部署安全性。
arXiv:2606.09388v1 Announce Type: new Abstract: Deploying safe large language models (LLMs) on resource-constrained edge devices presents a critical c…
基于结构信息理论的黑盒不确定性量化方法，无需访问模型内部，已被UAI 2026接收，为LLM可靠性评估提供新思路。
arXiv:2511.16275v4 Announce Type: replace-cross Abstract: Reliable uncertainty quantification (UQ) is essential for deploying large language models (L…