1
独家｜获超亿美元融资，Sand.ai 曹越：为什么视频是通往世界模型最重要的路径
Sand.ai获超亿美元融资，曹越详解为何视频是通往世界模型的关键路径
“每一代模型，我们都在押注一个非共识。” 文｜邓咏仪 编辑｜张雨忻 Sand.ai 创始人曹越，不太关心自己站在共识的哪一边。 Sand.ai 是一家视频生成模型和产品公司，成立于2024年1月。曹越创立Sand.ai 的故事也已经被讲过很多遍：在上一段创业“光年之外”戛然而止后，曹越很快就投入到 …
2
Logit Distance Bounds Representational Similarity
理论揭示：判别模型的条件分布等价性如何约束内部表示的唯一性，为理解模型表征相似性提供新视角
arXiv:2602.15438v3 Announce Type: replace-cross Abstract: For a broad family of discriminative models that includes autoregressive language models, id…
3
GPT-Based Fast Simulation of CLAS12 Detector Hits via Conditional Autoregressive Generation
用GPT做物理探测器模拟？论文提出条件自回归生成方法，加速CLAS12探测器响应仿真，AI与高能物理的跨界创新。
arXiv:2606.16035v1 Announce Type: cross Abstract: Modern particles physics experiments have demonstrated an increasing need for fast, high-fidelity de…
4
Google's latest DiffusionGemma open AI model comes with a 4x speed boost
Google开源DiffusionGemma，非自回归文本模型，速度提升4倍，类似图像去噪生成。
Diffusion AI is most common in image generation, but it can make text outputs much faster.
5
MAGE: All-[MASK] Block Already Knows Where to Look in Block Diffusion LLM
块扩散语言模型MAGE让所有[MASK]块自主定位关键位置，显著提升生成速度与连贯性。
arXiv:2602.14209v2 Announce Type: replace Abstract: Block diffusion LLMs are an emerging paradigm for parallel language generation, but their KV cachi…
6
When Autoregressive Consistency Hurts Safety Alignment
自回归一致性让大模型安全对齐脆弱不堪：微调只能重塑输出开头几个token，后续轨迹难以纠正。
arXiv:2606.04168v1 Announce Type: new Abstract: Safety alignment in large language models (LLMs) is fragile in part because it is often shallow: fine-…
7
Compositional Generalization in Autoregressive Models via Logit Composition
新方法通过logit组合提升自回归模型的组合泛化能力，突破结构泛化瓶颈。
arXiv:2605.28304v1 Announce Type: new Abstract: Composing autoregressive models remains a core challenge in understanding how large language models ca…
8
SONAR-LLM: Autoregressive Transformer that Thinks in Sentence Embeddings and Speaks in Tokens
打破逐字生成范式，在连续句子嵌入空间中“思考”，再解码为文本，革新了大模型推理方式。
arXiv:2508.05305v2 Announce Type: replace Abstract: The recently proposed Large Concept Model (LCM) generates text by predicting a sequence of sentenc…
9
MiniGPT: Rebuilding GPT from First Principles
从零复现GPT核心机制，基于PyTorch实现简洁自回归语言模型，AI学习者必读的底层论文教程。
arXiv:2605.17398v1 Announce Type: cross Abstract: This paper presents MiniGPT, a compact from-scratch implementation of GPT-style autoregressive langu…
10
Delta Forcing: Trust Region Steering for Interactive Autoregressive Video Generation
提出Delta Forcing方法，解决交互式自回归视频生成中响应性与稳定性的平衡难题。
arXiv:2605.14382v2 Announce Type: replace Abstract: Interactive real-time autoregressive video generation is essential for applications such as conten…
11
Matrix-Decoupled Concentration for Autoregressive Sequences: Dimension-Free Guarantees for Sparse Long-Context Rewards
自回归序列的矩阵解耦集中不等式，为稀疏长上下文奖励提供无维度保证，理论创新突破。
arXiv:2605.06017v2 Announce Type: replace Abstract: Sequence-level evaluations in autoregressive Large Language Models (LLMs) rely on highly dependent…
12
Conditional Attribute Estimation with Autoregressive Sequence Models
这篇论文提出了利用自回归序列模型进行条件属性估计的新方法，直击生成模型在全局结构控制上的痛点，值得关注。
arXiv:2605.14004v1 Announce Type: new Abstract: Generative models are often trained with a next-token prediction objective, yet many downstream applic…
13
Motion-Aware Caching for Efficient Autoregressive Video Generation
提出运动感知缓存复用策略，显著加速自回归视频生成过程。
arXiv:2605.01725v2 Announce Type: replace-cross Abstract: Autoregressive video generation paradigms offer theoretical promise for long video synthesis…
14
FlipAttack: Jailbreak LLMs via Flipping
揭秘LLM从左到右理解的弱点：仅需在左侧加噪声，就能轻松绕过黑盒大模型的安全护栏。
arXiv:2410.02832v2 Announce Type: replace-cross Abstract: This paper proposes a simple yet effective jailbreak attack named FlipAttack against black-b…
15
Representation Without Reward: A JEPA Audit for LLM Fine-Tuning
提出用JEPAs审计LLM微调：预测隐含表示而非输出，以提升任务指标。
arXiv:2605.15394v1 Announce Type: cross Abstract: Joint-embedding predictive architectures (JEPAs) propose that a model should learn more useful abstr…
16
Variational Autoregressive Networks with probability priors
结合概率先验的变分自回归网络，一种解决蒙特卡洛临界慢化的新方法
arXiv:2605.16020v1 Announce Type: new Abstract: Monte Carlo methods are essential across diverse scientific fields, yet their efficiency is frequently…