【赞助商】
OpenClaw快报
每天五分钟,听听 OpenClaw 快报,带你了解最新动态和业内讨论
传送门 www.xiaoyuzhoufm.com
【目录】
本期的 15 篇论文如下:
00:28 🎯 GroundingPI: A Grounding Foundation Model towards Physical Intelligence with Visual Primitives(GroundingPI:面向物理智能的视觉基元定位基础模型)
01:13 🤖 AgentSWE: Can Coding Agents Build the Agent You Actually Want?(AgentSWE:编码智能体能否构建你真正想要的智能体?)
01:58 🤖 Scaling Agentic Data for Long-Horizon Terminal Intelligence(面向长时程终端智能的智能体数据规模化)
02:45 🧠 Thinking Inertia: LLMs Keep Thinking When Told Not To(思维惯性:大语言模型被要求不思考时仍持续思考)
03:29 🎬 VIS-Ground: Video Interactive Storytelling with Contextual Grounding(VIS-Ground:基于上下文锚定的视频交互式叙事)
04:13 🔀 Multi-Stream Attention Residuals(多流注意力残差)
05:02 🎯 Relevance Does Not Imply Applicability: Experience Activation for Personal GUI Agents(相关性不等于适用性:面向个人 GUI 智能体的经验激活)
05:50 🏙 CrossTimeEdit: A Decade-Spanning Cross-View Dataset and Reward-Guided Editing for Historical Street-View Generation(CrossTimeEdit:面向历史街景生成的十年跨度跨视角数据集与奖励引导编辑)
06:33 ⚡ GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed(GroundAnything:以闪电速度协调并行解码与精确视觉定位)
07:13 🧠 SpatialASM: Internalizing Active Mentalization for Spatial Reasoning(SpatialASM:内化主动心智化以实现空间推理)
07:58 🤖 Systematic Multi-Agent Vision-and-Language Navigation: Formulation, Benchmark, and Method(系统化多智能体视觉与语言导航:形式化、基准与方法)
08:44 🌐 T3lescope: Arbitrary-Resolution High-Fidelity Generative Surface Reconstruction from Images(T3lescope:基于图像的任意分辨率高保真生成式表面重建)
09:26 🧠 Learning More by Training Less: Uncertainty-Guided Progressive Contraction for Efficient Reasoning Post-Training(以更少训练学习更多:面向高效推理后训练的不确定性引导渐进收缩)
10:12 🧩 ARCS: Towards Precise Text-to-SQL via Structured Disambiguation(ARCS:通过结构化消歧迈向精准的文本到SQL)
10:57 🤖 SPLATIFY: Reproduce, Discover, Innovate! From Papers and Ideas to Trainable 3DGS Code(SPLATIFY:复现、发现、创新!从论文与想法到可训练的3DGS代码)

【关注我们】
您还可以在以下平台找到我们,获得播客内容以外更多信息
小红书: AI速递
