arXiv 论文速递

Robots deployed in the physical world must be able to improve beyond their initial training as they encounter new situations and failures. For this improvement to scale across task…

The dynamics of generative models exhibit two apparently distinct temporal windows: a speciation window, in which a sample commits to a semantic class, and a nonlocality window, in…

Linear attention replaces growing KV caches with fixed-size recurrent states, yet these persistent states can become a substantial memory bottleneck under concurrent serving. Direc…

针对线性注意力中循环状态成为内存瓶颈的问题，提出精确的循环状态量化方法，使混合架构大模型在保持精度的同时显著降低内存开销。

Recent LLMs increasingly adopt hybrid designs that replace standard attention with linear attention, such as Gated DeltaNet (GDN) and Kimi Delta Attention (KDA). Although they comp…

The landscape of satellite imagery time series datasets and boundary-pushing architectures for cropland segmentation has never been richer. However, in this gold rush, important tr…

Evaluations of graph reconstruction by language models typically report a single aggregate distance between the original and the reconstructed graph. We prove that for the Wasserst…

提出以实体传记替代纯时间线描述来增强长视频记忆，解决跨小时连接同一对象事件的难题，对长视频问答与记忆建模有重要价值。

Answering questions about long videos often requires connecting events involving the same objects across hours or days. Chronological descriptions and text-derived entities can lea…

提出元推理机制，让智能体在执行前先规划推理路径，解决长程任务中执行控制随步骤增加而失控的问题，对提升智能体推理可扩展性有重要意义。

As agents take on longer and more complex problems, controlling the execution becomes a task in its own right. Each step in the run brings new control choices, like which partial w…

Agent performance depends on both reasoning ability and the environment in which it acts. We study test-time AI-for-AI, asking how a Builder can learn to construct better execution…

训练小型顾问模型通过自然语言建议引导冻结的前沿大模型执行器，结合任务奖励与多轮自蒸馏，为低成本提升大模型能力提供新路径。

A small trainable advisor can steer a frozen language-model executor using natural-language advice. In addition to learning from task rewards, the advisor can use feedback from com…

指出传统MoE在视频扩散模型中存在专家分配均匀化陷阱，提出SplitMoE实现更有效的专家路由，为视觉生成模型的规模化提供新范式。

Mixture-of-Experts (MoE), popularized by large language models, is a promising paradigm for scaling visual generative models. However, conventional token-wise MoE routes tokens ind…

Generative modeling is widely used for producing diverse objects from complex, multimodal distributions. However, its expressivity does not, in general, come with formal guarantees…