👍 120
09/08 08:00
In this work, we present Gander, an end-to-end model that unifies omni perception, realtime interaction, and agentic capabilities within a single framework. In contrast to turn-based conventional paradigms, Gander continuously receives streaming inputs across multiple modalities, including video, sp
中文介绍 提出端到端模型 Gander,在单一框架内统一全模态感知、实时交互与智能体能力。与传统轮次式范式不同,它持续接收视频、语音等流式多模态输入,实现连续理解与响应,将感知、交互和行动无缝衔接,为实时多模态智能体提供一体化方案。
👍 55
09/07 08:00
World-Action Models inherit world knowledge from video-generative priors, and channel it into executable control signals through embodied experience. Existing systems, however, are monolithic: the generative backbone, visual representation, architecture, information flow, inference procedure, and tr
中文介绍 针对世界-动作模型(WAM)系统多为单体式、各组件难以替换的问题,提出开源模块化框架 OpenWAM,将生成主干、视觉表示、架构、信息流与推理流程解耦,系统探索 WAM 预训练与扩展规律,为可复现的世界-动作模型预训练提供开放基线。
👍 49
09/04 08:00
World-action models (WAM) predict future states to guide robot actions, enabling learning from both action-free video and action-labeled interaction. Most inherit pretrained video generators, leaving WAM pretraining and scaling underexplored. We introduce Genie Envisioner Act 2.0 (GE-Act 2.0), a wor
中文介绍 世界-动作模型多直接沿用预训练视频生成器,其自身预训练与规模扩展仍缺研究。提出 GE-Act 2.0,在 Genie Envisioner 基础上进行世界-动作模型预训练与缩放,兼顾无动作视频与带动作交互数据,预测未来状态以指导机器人操作,推动机器人操控策略学习。
👍 46
09/08 08:00
We present Miles v0.1, a full-stack, production-ready system for frontier post-training. Building upon the clean design of slime, Miles designs each stage of the reinforcement-learning (RL) training loop around a single principle: components should be verified, clean, and customizable. With accuracy
中文介绍 提出生产级后训练全栈系统 Miles v0.1,基于 slime 的简洁设计,围绕「组件可验证、干净、可定制」原则构建强化学习训练循环各阶段,面向前沿模型的后训练需求,兼顾精度与工程可用性,为 RL 后训练提供可直接落地的完整方案。
👍 43
09/08 08:00
Autoregressive (AR) video diffusion models have shown great potential in real-time video generation. Recent methods distill pretrained bidirectional video diffusion models into causal AR students through Distribution Matching Distillation (DMD), but the generated videos often suffer from over-satura
中文介绍 自回归视频扩散蒸馏常因 DMD 训练出现画面过饱和等问题。提出 Mask Forcing,通过双噪声掩码 rollout 改进蒸馏:对输入施加双路噪声掩码,抑制误差累积与分布偏移,从而在保持实时生成的同时提升自回归视频蒸馏的视觉质量。
👍 32
09/04 08:00
Large Reasoning Models (LRMs) achieve superior problem-solving through extended Chain-of-Thought (CoT) generation, but the resulting key-value (KV) cache grows linearly with sequence length and creates severe memory bottlenecks, often exceeding GPU capacity for long reasoning traces. Existing KV cac
中文介绍 大型推理模型的长 CoT 使 KV cache 随序列线性增长,成为显存瓶颈。提出 BeaconKV,利用 beacon 查询引导 KV cache 压缩,识别并保留对后续推理关键的 token,在长推理场景下大幅降低显存占用,同时维持推理性能。
👍 30
09/06 08:00
Latent visual reasoning aims to perform multimodal reasoning through hidden-state computation rather than explicit textual chains of thought. However, visual information being present in a latent state does not imply that the model actually relies on that state when producing its answer, especially
中文介绍 潜空间视觉推理通过隐状态而非显式文本链进行多模态推理,但视觉信息存在于隐状态并不意味着模型真正依赖它作答。论文对此展开分析并提出方法,使潜空间视觉推理成为必要,验证模型确实利用隐藏视觉表征完成推理。
👍 27
09/09 08:00
Foundation vision-language models (VLMs) exhibit broad intelligence about the world, yet translating this intelligence into robot control remains challenging. We present Show-Harness, an Embodied Harness that enables VLMs to "play" robots through a compact semantic interface linking intent to action
中文介绍 基础 VLM 具备广泛世界知识,但转化为机器人控制仍困难。提出 Show-Harness,一种具身 harness,通过紧凑语义接口把意图映射为动作,让 VLM 智能体直接「玩」机器人,降低对专用策略网络的依赖,实现通用 VLM 的机器人操控。
👍 27
09/05 08:00
As large language models (LLMs) are increasingly deployed in alignment-sensitive contexts, activation steering has emerged as a lightweight, inference-time alternative to fine-tuning methods (e.g., RLHF, DPO) for behavioral control. However, existing work typically validates steering on isolated beh
中文介绍 激活引导是替代 RLHF、DPO 的轻量推理期行为控制手段,但以往验证多限于孤立行为。论文在 LLM 引导空间中检验人类价值几何结构,系统验证引导方向与价值维度的对应关系,为基于激活引导的可控对齐提供几何层面的证据。
👍 25
09/07 08:00
Transferring human hand demonstrations to robotic grippers has recently emerged as a cost-effective solution for robot learning. However, existing methods are largely confined to simple, planar tasks and fail to handle complex spatial movements (e.g., intricate trajectories involving rotations or fl
中文介绍 人手演示迁移到机器人夹爪是低成本学习方案,但现有方法多限平面简单任务,难以处理含旋转等复杂空间运动。论文发布 CosmoH2G 手到夹爪迁移数据集,并提供基线方法,面向复杂空间轨迹下的物体操作,推动该方向的数据与算法研究。
👍 23
09/08 08:00
Large language models are increasingly deployed as agents that plan over long horizons and act through external tools. Most agents select actions through unconstrained generation over an accumulating history, leaving implicit the procedural knowledge of what to do, in what order, and under which con
中文介绍 LLM 智能体多在累积历史中无约束生成动作,隐含了「做什么、按什么顺序、在何种条件下做」的程序性知识。提出 Procedural Graphs,用自演化执行结构显式化并复用这类知识,让智能体在长程规划与工具调用中更稳定、可解释。
👍 21
09/05 08:00
Multimodal Large Language Models (MLLMs) perform strongly on general visual understanding tasks such as visual question answering, yet they often struggle with a basic comparative skill: identifying what has changed between two similar images. We introduce VDiff-Bench, a challenging multiple-choice
中文介绍 多模态大模型在通用视觉问答上表现强,却难以完成识别两张相似图像差异这一基础比较能力。提出 VDiff-Bench,一个细粒度图像差异识别的多选题基准,用于评测 MLLM 对细微视觉变化的辨别能力,暴露现有模型在该任务上的不足。
👍 21
09/04 08:00
Vision-Language-Action (VLA) models have shown promising progress in language-conditioned robotic manipulation. However, existing datasets and benchmarks mainly evaluate task completion under predefined settings, offering limited insight into model reasoning under increasing spatial and procedural c
中文介绍 VLA 模型在语言条件机器人操控上进展明显,但现有数据集与基准多在预设场景下评测任务完成度,难以反映空间与流程复杂度上升时的推理能力。提出 RoboSPA 基准,考察 VLA 能否超越简单场景与短时程任务,系统评估其复杂环境泛化能力。
👍 20
09/08 08:00
Representing a 3D scene as multi-view images allows 2D VLMs to reason in 3D by reusing priors from pre-training, sidestepping the scarcity of annotated 3D data. However, it produces thousands of redundant visual tokens whose cost grows with every view. Existing visual token pruners fall into two fam
中文介绍 以多视图图像表示 3D 场景可让 2D VLM 复用预训练先验进行 3D 推理,但视觉 token 随视角增加而冗余膨胀。提出 CoVeR,基于覆盖度的 token 剪枝方法,在保留 3D 推理所需信息的前提下压缩多视图 token,降低计算开销。
👍 18
09/09 08:00
Recent video world models generate increasingly realistic and interactive visual experiences, yet lack reliable mechanisms for maintaining persistent world state and enforcing programmable rules over extended interactions. We introduce Programmable World Model, a framework that decouples world-state
中文介绍 视频世界模型能生成逼真交互画面,却缺乏在长程交互中维持持久世界状态与可编程规则的机制。提出 Programmable World Model,将世界状态与渲染解耦,允许以规则编程方式约束世界演化,从而支持可持久、可编程的交互式世界模拟。
👍 18
09/06 08:00
Diffusion-based models enable monocular geometry estimation, yet their pixel-space precision is limited by a shared, under-studied error source: VAE reconstruction degradation. The 8x spatial compression in the VAE encoder-decoder degrades surface normals at object boundaries; even encoding and deco
中文介绍 扩散模型用于单目几何估计时,精度受 VAE 编解码 8 倍空间压缩导致的边界法向退化制约。提出 TransNormal-2,采用几何约束的 rectified flow 与边缘感知解码,缓解 VAE 重建退化,提升物体边界处表面法向估计的像素级精度。
👍 16
09/05 08:00
We present Cadence, an error-bounded lossy compressor for numeric time series pairing a 330M-parameter time-series foundation model (Google TimesFM-3) with an adaptive arithmetic coder, guaranteeing |x_t-x_t|leτ on every sample. One negative result constrains the design space: for lossless coding a
中文介绍 提出 Cadence,面向数值时间序列的误差有界有损压缩器,将 330M 参数时序基础模型 TimesFM-3 与自适应算术编码结合,保证每个样本满足 |x̂_t−x_t|≤τ。同时报告一个负结果,界定了无损编码条件下的设计空间约束。
👍 15
09/07 08:00
In this paper, we propose ReactVAU, a Slow-Fast Decoupled Framework for real-time streaming Video Anomaly Understanding (VAU). Existing VAU methods rely on offline inference with global temporal sampling, which violates causality and prevents deployment in live surveillance streams. Conversely, gene
中文介绍 现有视频异常理解依赖离线全局采样,违背因果性,难以部署于实时监控流。提出 ReactVAU 慢-快解耦框架,用慢分支理解语义、快分支处理流式输入,实现实时流式视频异常理解,兼顾因果推理与响应速度。
👍 15
08/29 08:00
Large language models often answer structurally unanswerable questions, such as computing cot(-540°) or evaluating (1).startswith("1"), instead of abstaining. We ask whether this failure reflects missing recognition or failed routing from recognition to abstention. Across instruction-tuned models fr
中文介绍 LLM 常回答结构上不可回答的问题,如计算 cot(-540°) 或判断 (1).startswith("1"),而非拒答。论文在多个指令微调模型上探究该失败源于「未识别不可回答」还是「识别后未路由到拒答」,提出识别-拒答错位概念并分析其成因。
👍 11
09/08 01:56
Chain-of-Thought (CoT) improves the reasoning ability of Large Language Models (LLMs) but incurs substantial computation and context costs. Existing methods either lose intermediate information through hard pruning or lack a principled criterion for continuous compression. We present A*-Thought-V2,
中文介绍 CoT 提升 LLM 推理能力却带来计算与上下文开销。提出 A*-Thought-V2,利用 LLM 的几何动力学进行高效潜空间推理,以有原则的连续压缩替代硬剪枝,保留中间信息并减少 token 消耗,在降低推理成本的同时保持推理性能。