从理论层面重新审视LLM对齐的奖励分布假设，揭示重尾效应对对齐边界的影响，值得关注。
arXiv:2604.10727v2 Announce Type: replace-cross Abstract: Classical information-theoretic learning bounds typically rely on KL mutual information and …
重尾数据下PCA的经典方法失效？这篇论文提出了一种鲁棒主成分分析方法，突破二阶矩限制。
arXiv:2603.11308v3 Announce Type: replace Abstract: Principal Component Analysis (PCA) is a cornerstone of dimensionality reduction, yet its classical…
当奖励分布出现“重尾”异常值，批量决策的赌博机问题如何逼近最优？全新算法与理论分析一次讲透。
arXiv:2510.03798v3 Announce Type: replace Abstract: The batched multi-armed bandit (MAB) problem, where rewards are collected in batches, is pivotal i…
重尾分布下的扩散模型新突破：自调节退火机制，无需手动调参即可优化生成质量。
arXiv:2606.01645v1 Announce Type: cross Abstract: Diffusion models have emerged as a leading framework for deep generative modeling. While the standar…
突破传统统一学习率，重尾分布指导LLM逐层自适应学习，大幅提升训练效率与模型性能。
arXiv:2605.22297v1 Announce Type: cross Abstract: Learning rate configuration is a fundamental aspect of modern deep learning. The prevailing practice…
提出首个在重尾MDP上同时实现随机与对抗环境最优遗憾的BoBW算法，突破保守局限。
arXiv:2602.01295v3 Announce Type: replace Abstract: We investigate episodic Markov Decision Processes with heavy-tailed losses (HTMDPs). Existing appr…