HuggingFace 每日AI论文速递

HuggingFace 每日AI论文速递 的全部节目

共 747 集。

2026.11.18 | 国产栈训练Xing4;内核优化也算成本

2026.11.18 | 国产栈训练Xing4;内核优化也算成本

HuggingFace 每日AI论文速递

⏱️ Highlights [00:29] 🧩 TeleChat to Xing4: Advancing Frontier Agentic Models on Domestic Chips and Frameworks(从TeleChat到Xing4:在国产芯片与框架上推进前沿智能体模型) [01:11] ⚙️ AscendGYM: A Cost-Aware Kernel Optimization Benchmark for Real-World Workloads on Ascend NPUs(AscendGYM:面向昇腾NPU真实负载的成本感知内核优化基准) [01:50] 📝 Weight-Space Programming Doesn't Need Transformers(权重空间编程不需要Transformer) [02:29] 🦾 ForceHarness: Turning Force into Action and Physics for Frozen Frontier Models(ForceHarness:为冻结的前沿模型把力转化为动作与物理信息) [03:11] 🩺 Breath-Dx: A Time Series Classification Benchmark for Disease Screening with eNose(Breath-Dx:使用电子鼻进行疾病筛查的时间序列分类基准) [03:55] 🌿 Multi-Branch Transformer(多分支Transformer) [04:36] 🧠 Self-Improvement Memory Enables Continual Recursive Self-Improvement(自我改进记忆使持续递归自我改进成为可能) [05:15] 🎯 TCP4D: A Unified Tool-Center-Point Tracking and Generation Model for Robot Manipulation(TCP4D:面向机器人操作的统一工具中心点跟踪与生成模型) [05:55] 🎬 FastFrames: Video Generation at Interactive Latency(FastFrames:实现交互级延迟的视频生成) [06:37] 🔍 VideoNexus: Video Perception Beyond the Reasoning Context(VideoNexus:超越推理上下文的视频感知) [07:16] 🤝 Heave-Ho: Decentralized Long-Horizon Multi-Humanoid Collaboration(Heave-Ho:去中心化的长时程多人形机器人协作) [07:55] 🗜️ FMT: Lossless Image Compression in a Fixed Number of Parallel Steps(FMT:以固定数量并行步骤实现无损图像压缩) [08:36] 🏥 MedJev: A System One Vision-Language Decision Model for Fast and Effective Medical Image Understanding(MedJev:用于快速有效医学图像理解的系统一视觉语言决策模型) [09:19] 🖼️ MPEG-Diffusion: Mixed-Resolution Video Generation(MPEG-Diffusion:混合分辨率视频生成) [09:58] 🔄 Agentic On-Policy Distillation Implicitly Reshapes LLMs' ReAct Patterns(智能体同策略蒸馏会隐式重塑大语言模型的ReAct模式) [播客论文目录配图] 【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

11 分钟 21 次播放 5小时前

2026.11.17 | 开放模型全流程;联合生成交易仓位

2026.11.17 | 开放模型全流程;联合生成交易仓位

HuggingFace 每日AI论文速递

⏱️ Highlights [00:29] 🌐 K2-Horizon: A Frontier Fleet of Open Source Language Models(K2-Horizon:前沿开源语言模型系列) [01:07] 📈 AllocFlow: End-to-End Portfolio Trading Action Generation from Raw Market Data(AllocFlow:从原始市场数据端到端生成投资组合交易动作) [01:45] 🔬 Sage S2.1 Flash Technical Report(Sage S2.1 Flash 技术报告) [02:23] 🩺 InclusionMed: Benchmarking AI for Inclusive Healthcare(InclusionMed:面向普惠医疗的AI基准评测) [03:03] 🛡️ SPILLWAY: Separating Target Admission from Sanitization to Defend Tool-Using Agents against Prompt Injection(SPILLWAY:分离目标准入与清洗,防御工具使用智能体的提示注入) [03:42] 💻 SWE-CoWork: Benchmarking Coding Agents as Digital Employees in Organizations(SWE-CoWork:将编码智能体作为组织中的数字员工进行评测) [04:21] 📝 PaperCritic: Training AI Reviewers to Assess the Soundness of Papers(PaperCritic:训练AI评审员评估论文的可靠性) [05:00] 🎙️ VoxMel TTS Technical Report(VoxMel 文本转语音技术报告) [05:38] 🤖 ScaleVLA: Scaling Vision-Language-Action Models to 552 Billion Parameters(ScaleVLA:将视觉—语言—动作模型扩展至5520亿参数) [06:16] ⚙️ Accelerate Bench: Evaluating Agents for Application-Level Program Optimization(Accelerate Bench:评估智能体的应用级程序优化能力) [06:56] 🧬 Calpha: Toward Native Biomolecular Intelligence in a General-Purpose Language Model(Calpha:迈向通用语言模型中的原生生物分子智能) [07:34] 🦾 SALA: Semantically Aligned Latent Actions Enable Dual-Axis Scaling of Embodied Foundation Models(SALA:语义对齐的潜在动作实现具身基础模型的双轴扩展) [08:12] 🎬 MovieRoll: A Systematic Framework for Rolling Out Long-Form Videos(MovieRoll:用于逐步生成长视频的系统性框架) [08:52] 🔁 Latent Feedback Is a Compute and Data Amplifier(潜在反馈是计算与数据的放大器) [09:29] 🚀 LoopFill: Pipelined Multi-Token Speculative Decoding for Looped Transformers(LoopFill:面向循环Transformer的流水线多词元推测解码) 【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

10 分钟 42 次播放 1天前

2026.11.16 | 机器人如何验证行动;编程智能体搭建动态场景

2026.11.16 | 机器人如何验证行动;编程智能体搭建动态场景

HuggingFace 每日AI论文速递

⏱️ Highlights [00:30] 🤖 Envisioning Agentic Robotics: A New Era of Robotic Systems(智能体机器人学展望:机器人系统的新时代) [01:04] 🎬 VibePrevisBench: Evaluating Agentic Coding for 4D Scene Generation in Blender(VibePrevisBench:评估用于 Blender 四维场景生成的智能体编程) [01:41] 🎮 Gamer: A Generalist Vision-Language-Action Foundation Model for Long-Horizon Real-Time Game Agent(Gamer:面向长时程实时游戏智能体的通用视觉—语言—动作基础模型) [02:19] ⚡ MMFlash: Faster Speculative Decoding for Vision-Language Models(MMFlash:面向视觉语言模型的更快推测解码) [02:55] 🎞️ V-JEPA-LM: Video-Language Models That Read Dynamics Beyond Static, Text-Aligned Representations(V-JEPA-LM:超越静态文本对齐表征、读懂动态的视频语言模型) [03:33] 🔁 Loopback Transformer: Towards a Unified View of Residual Looped Transformers(Loopback Transformer:迈向残差循环 Transformer 的统一视角) [04:10] 🧠 MIRIX-2.0: Multi-Agent Memory System(MIRIX-2.0:多智能体记忆系统) [04:51] 📹 JVT: Just Video Transformer makes pixel-space video diffusion simple(JVT:Just Video Transformer 让像素空间视频扩散更简单) [05:29] 🔍 Rethinking in Spikes: Mitigating Hallucinations in MDLMs with Step-Aware Decoding(在尖峰处重新思考:用步骤感知解码缓解多模态扩散语言模型的幻觉) [06:08] 🗃️ MEM-SQL: Memory-Empowered Text-to-SQL(MEM-SQL:记忆增强的文本到SQL) [06:46] 🤝 Aligning AI Teams(对齐AI团队) [07:24] 🧬 Emergent Biological Capabilities in a Foundation Model for Molecular Interactions(分子相互作用基础模型中涌现的生物学能力) [08:03] 🚘 CANTraj: Recovering Vehicle Trajectories from CAN Signals in Naturalistic Driving(CANTraj:从自然驾驶中的CAN信号恢复车辆轨迹) [08:43] 🎥 Disen-Forcing: Disentangling Semantic Anchoring from Motion for Autoregressive Video Diffusion(Disen-Forcing:为自回归视频扩散解耦语义锚定与运动) [09:22] 📋 GH-AWTrace: A Large-Scale Dataset of GitHub Agentic Workflow Execution Logs(GH-AWTrace:GitHub智能体工作流执行日志大规模数据集) 【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

10 分钟 49 次播放 2天前

2026.11.13 | 快速决策也要可靠;双视角驾驶世界模型

2026.11.13 | 快速决策也要可靠;双视角驾驶世界模型

HuggingFace 每日AI论文速递

⏱️ Highlights [00:24] ⚖️ SO-Arena: Benchmarking the Capability Spectrum and Reliability of System One Models(SO-Arena:评测系统一模型的能力谱系与可靠性) [01:00] 🚗 BirdDreams: Joint Perspective and BEV Video Generation for a Plausible, Reactive Driving World Model(BirdDreams:联合生成透视与鸟瞰视频,构建合理且可响应的驾驶世界模型) [01:34] 🧩 G-STAR: Group-Scaled Token Capacity for Deeply Compressed Autoregressive Generation(G-STAR:以分组扩展词元容量,实现深度压缩的自回归生成) [02:08] 🎙️ Duo: A Reciprocal World Model for Full-Duplex Interaction(Duo:面向全双工交互的互惠世界模型) [02:43] 🎬 PhysLoop: Harnessing LLMs' Physics Awareness for Self-Improving Video Generation Models(PhysLoop:利用大语言模型的物理认知,支持视频生成模型自我改进) [03:19] 🔗 On the Design of Hyper-Connections: Simplification, Stability, and Scalability(论超连接的设计:简化、稳定性与可扩展性) [03:55] 🧑‍🏫 Actor, Judge, or Teacher? Understanding Decision Models in Long-Horizon Agents(行动者、裁判还是教师?理解长时程智能体中的决策模型) [04:29] 📷 TTB: Test-time Baking for Efficient Rendering of Decoder-only View Synthesis Models(TTB:通过测试时烘焙,高效渲染纯解码器视角合成模型) [05:07] 🦾 Ropedia Xperience 2: A Gold Standard for Scaling Physical AI Environments and Behaviors(Ropedia Xperience 2:规模化扩展物理 AI 环境与行为的黄金标准) [05:46] 🎨 Open Vision Banana: A Fully Open Recipe for Generative Visual Perception(Open Vision Banana:生成式视觉感知的完全开放配方) [06:21] 🌐 WorldDraft: If Code Builds the World, Can Video Models Just Render It?(WorldDraft:如果代码构建世界,视频模型能只负责渲染吗?) [06:57] 💾 Factored One-sided Adam-Moment for Practical and Scalable SOAP(面向实用且可扩展 SOAP 的分解式单侧 Adam 矩) [07:33] ✨ UniSpark: Exploring Hybrid Attention for Multimodal Understanding and Generation(UniSpark:探索用于多模态理解与生成的混合注意力) [08:08] 🛡️ Evidence-Guided Recovery for AI Agents with Ambiguous Tool Outcomes and Partial Execution(面向工具结果不明与部分执行的 AI 智能体的证据引导恢复) [08:46] 📁 Agent-R^2: Scaling Agents for Realistic Productivity Work via Real-Source Data Synthesis(Agent-R^2:通过真实来源数据合成,扩展智能体的现实生产力工作能力) 【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

9 分钟 43 次播放 5天前

2026.11.12 | 教材变证明训练场;量化智能体闭环纠错

2026.11.12 | 教材变证明训练场;量化智能体闭环纠错

HuggingFace 每日AI论文速递

⏱️ Highlights [00:29] 📚 Scaling Mathematical Proof Supervision from Structured Textbooks(从结构化教材扩展数学证明监督) [01:03] 🧭 SteerDistill: Quantizing LLM Agents via Closed-Loop Teacher Steering(SteerDistill:通过闭环教师引导量化大语言模型智能体) [01:40] 📖 Incentivizing Self-Learning in Large Language Models via Reinforcement Learning(通过强化学习激励大语言模型自主学习) [02:17] 🎮 GTAGym: Real AAA Game Tasks Environment for Evaluation and Reinforcement Learning(GTAGym:用于评估与强化学习的真实 AAA 游戏任务环境) [02:51] 🔎 Read Less, Solve More: Token-Efficient Sparse Reading of External Content for AI Agents(少读多解:AI 智能体对外部内容的高效 token 稀疏阅读) [03:28] 🔀 Ensemble Routing for Prompt-Adaptive Cost-Aware LLM Inference(面向提示自适应、成本感知大语言模型推理的集成路由) [04:04] 🪞 Where Does Alignment Emerge and What Does It Mean? A Closer Look at the Platonic Representation Hypothesis(对齐在哪里涌现,又意味着什么?细察柏拉图表征假说) [04:41] 🧠 WMIP: A Working-Memory-Inspired Programmable Probe of Cumulative State Tracking in Large Language Models(WMIP:受工作记忆启发、探测大语言模型累积状态追踪的可编程工具) [05:20] 🧪 CogBattery: Human-Referenced Evaluation of Cognitive Task Behavior and Structure in Language Models(CogBattery:以人类为参照评估语言模型的认知任务行为与结构) [05:56] 🕸️ Module-Aware Optimization for Graph Neural Networks(图神经网络的模块感知优化) [06:32] 🫀 Mixed Signals: Weight-Space Programming for EEG and ECG(Mixed Signals:面向 EEG 与 ECG 的权重空间编程) [07:10] 👁️ Interactive Visual Gym: Evaluating Visual Interaction Models with Simulated Users in Interactive Worlds(Interactive Visual Gym:在交互世界中用模拟用户评估视觉交互模型) [07:47] ⚙️ Evolving Infra: Agent-Adaptive LLM Serving at Scale(Evolving Infra:大规模、智能体自适应的大语言模型服务) [08:26] 🎚️ Precision on Demand: Escalating Weight Bits at Uncertain Decoding Steps in a Small Local Model(按需精度:在小型本地模型的不确定解码步骤提升权重位宽) [09:00] 🛒 CARTBench: Rubric-Based Evaluation of Shopping Advice for Complex and Ambiguous Requests with Timelines(CARTBench:基于评分准则评估带时间线的复杂、模糊购物请求建议) [播客论文目录配图] 【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

10 分钟 68 次播放 6天前

2026.11.11 | 给记忆留证据;通用离散视觉码

2026.11.11 | 给记忆留证据;通用离散视觉码

HuggingFace 每日AI论文速递

⏱️ Highlights [00:25] 🧠 Don't Just Trust Your Memory: Evidence-Linked Memory Reduces Memory-Induced Degradation in LLM Agents(别只相信你的记忆:证据关联记忆减少大语言模型智能体中记忆引发的性能退化) [00:59] 🧩 dNaViT: Discrete Codes are General-Purpose Visual Representations(dNaViT:离散码是通用视觉表示) [01:32] 🤖 World In-Context Learning for Agentic Robot Manipulation(面向智能体式机器人操作的世界上下文学习) [02:07] 🖐️ DexForesight: Grounding Dexterous Policies with Implicit Spatial Foresight(DexForesight:以隐式空间预见能力为灵巧操作策略奠基) [02:44] 🎯 MMDM: Measuring Massive Multitask Decision Making(MMDM:衡量大规模多任务决策能力) [03:20] 🎬 Few-Step Streaming Video Generation via Condition Shifting(通过条件移位实现少步流式视频生成) [03:55] 🔧 Tree-Ising: Repairing Language Models with Bit-Level Ising Edits(Tree-Ising:通过比特级伊辛编辑修复语言模型) [04:33] 🌿 DepthGraft: Structural Regularization through Hierarchical Cross-Layer KV Reconstruction(DepthGraft:通过分层跨层KV重构实现结构正则化) [05:08] ⚙️ PhysOPSD: Code as Privilege for Physically Plausible Video Generation(PhysOPSD:以代码作为特权信息,生成物理合理的视频) [05:43] 🧬 ProTEUS: A mutation-resolved world model for closed-loop protein evolution(ProTEUS:用于闭环蛋白质进化的突变分辨世界模型) [06:19] 🎙️ WaveGen: Truly End-to-End Waveform Generation via Internal Spectral Trajectory Learning(WaveGen:通过内部频谱轨迹学习实现真正端到端的波形生成) [06:52] ⏰ Is Your Robot's Memory Just a Clock? Temporal Shortcuts in Memory-Augmented VLA Policies(你的机器人记忆只是个时钟吗?记忆增强VLA策略中的时间捷径) [07:28] 🛰️ EarthRecord: From Pixels to Land Change with Multimodal Agents over Four Decades of Satellite Imagery(EarthRecord:多模态智能体如何利用四十年卫星影像,从像素走向土地变化分析) [08:00] 🦾 ExRL: Sample-Efficient Online Adaptation of Long-Chunk Policies via Execution Length Control(ExRL:通过执行长度控制,实现长动作块策略的高样本效率在线适应) [08:34] 📈 Market-Conditioned JEPA: Learning from Unpredictable Market Shocks(市场条件化JEPA:从不可预测的市场冲击中学习) [播客论文目录配图] 【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

9 分钟 49 次播放 1周前

2026.11.10 | 让异构触觉共享经验;世界模型的统一工具箱

2026.11.10 | 让异构触觉共享经验;世界模型的统一工具箱

HuggingFace 每日AI论文速递

⏱️ Highlights [00:32] 🖐️ Learning Heterogeneous Tactile Representation for General Dexterous Manipulation(面向通用灵巧操作的异构触觉表征学习) [01:10] 🌍 WorldFoundry: Towards Unified Infrastructure for World Intelligence(WorldFoundry:迈向世界智能的统一基础设施) [01:50] 🔬 Sage-Taiyi: A Metacognitive Architecture for Scientific Discovery(Sage-Taiyi:面向科学发现的元认知架构) [02:27] 👁️ Reframe: Observation-Aligned Multi-Teacher On-Policy Distillation for Agentic Visual Reasoning(Reframe:面向智能体视觉推理的观察对齐多教师同策略蒸馏) [03:02] 🌿 Flide: Accelerating Decision Inference with Exact State Forking(Flide:通过精确状态分叉加速决策推理) [03:38] 🌉 ReproBridge: Building Reproducibility into ML Research Artifacts(ReproBridge:将可复现性融入机器学习研究产物) [04:13] 🧪 SIRIS: Can LLMs Tell When Simulation Errors Matter in AI-Driven Systems Research(SIRIS:大模型能否判断AI驱动的系统研究中仿真误差何时重要) [04:47] ⚖️ How Much Training-Inference Mismatch Can On-Policy Distillation Tolerate?(同策略蒸馏能容忍多大的训练—推理失配?) [05:26] 🔎 TRIO: Task Refinement with Individual Optimization for Retrieval-Augmented Generation(TRIO:面向检索增强生成的任务细化与独立优化) [06:02] 🎨 FewStepRL: Fine-tuning Few-step Diffusion Models with Reinforcement Learning(FewStepRL:用强化学习微调少步扩散模型) [06:40] 🛠️ InfraRSI: Harnessing Recursive Self-Improvement for Training Infrastructure(InfraRSI:利用递归自我改进优化训练基础设施) [07:19] 🏪 AutoVend: Benchmarking LLM Agents on Long-Horizon Autonomous Vending Operations(AutoVend:评估大模型智能体的长期自主售货运营) [07:57] 🧊 JoVox: Joint Hierarchical Flow Matching for Lightweight Voxel Generation(JoVox:用于轻量化体素生成的联合分层流匹配) [08:31] 🎯 Training Vision Transformers to Focus: Emergence of Spatial Attention Head Specialization(训练视觉Transformer学会聚焦:空间注意力头专门化的涌现) [09:08] 🤖 Auriga Technical Report: A Task-Agnostic Agent Harness for Zero-Shot Embodied Manipulation(Auriga技术报告:用于零样本具身操作的任务无关智能体运行框架) [播客论文目录配图] 【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

10 分钟 79 次播放 1周前

2026.11.09 | 高斯间距校准跨场景重建;参数先验加速多教师蒸馏

2026.11.09 | 高斯间距校准跨场景重建;参数先验加速多教师蒸馏

HuggingFace 每日AI论文速递

⏱️ Highlights [00:25] 🌐 SplatAnything: Gaussian Spacing Normalization for Feed-forward 3DGS across Arbitrary Scenes(SplatAnything:面向任意场景前馈式3DGS的高斯间距归一化) [01:02] 🧠 P^2-MOPD: Accelerating Multi-Teacher On-Policy Distillation via Parameter Priors(P²-MOPD:通过参数先验加速多教师在策略蒸馏) [01:43] 🎨 LongBanana: An Expert-Verified Benchmark for Long-Context Multi-Reference Image Synthesis(LongBanana:经专家核验的长上下文多参考图像合成基准) [02:21] 🔍 Beneath the Surface: What We Overlook When We Talk About Agent Evolution(表象之下:谈论智能体进化时,我们忽略了什么) [03:00] 🎬 EvoWorld: Towards Out-of-Sight World Evolution in Generative Video Models(EvoWorld:迈向生成式视频模型中的画外世界演化) [03:39] ⚙️ Fast and Exact Relative-Position Attention via Structure-Aware Kernels(通过结构感知内核实现快速且精确的相对位置注意力) [04:17] 🎙️ Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription(Confucius4-R2T2:面向真正实时转写的稳定前缀流式语音识别模型) [04:57] 🤝 AgentBaton: Efficient Model Collaboration in Agentic Tasks(AgentBaton:智能体任务中的高效模型协作) [05:37] 💻 Video-Coder: Reinforcing Agentic Coding for Video Generation(Video-Coder:通过强化智能体编码实现视频生成) [06:17] 🧩 Exploring Weight-Space Programming: Capabilities, Generalization, and Limitations(探索权重空间编程:能力、泛化与局限) [06:57] 🏗️ Frontier3DBench: Benchmarking Coding Agents for 3D Creation across Objects, Scenes, and Interactive Open Worlds(Frontier3DBench:跨物体、场景与交互式开放世界的编码智能体三维创作评测) [07:41] 📷 PREEN: Exemplar-Based Color Style Transfer with Smartphone Editor Presets(PREEN:使用智能手机编辑器预设的示例驱动色彩风格迁移) [08:18] 📊 LLMs Don't Know What They Know: Benchmarking Temporal Self-Knowledge for Financial Backtesting(大语言模型不知道自己知道什么:面向金融回测的时间自知能力评测) [09:00] 🔧 AutoCompress: Automated Hardware-Guided Model Compression(AutoCompress:自动化硬件引导模型压缩) [09:41] 🧱 OmniLatent: Unified Latent for 3D Part Modeling(OmniLatent:用于三维部件建模的统一潜表示) 【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

10 分钟 48 次播放 1周前

2026.11.06 | 视觉规划整块解码;借兄弟框架练智能体

2026.11.06 | 视觉规划整块解码;借兄弟框架练智能体

HuggingFace 每日AI论文速递

⏱️ Highlights [00:22] 🧭 VFlash: Fast, Accurate Visual Planning via Flash Primitive Decoding(VFlash:通过闪速图元解码实现快速、准确的视觉规划) [00:54] 🛠️ Harness-S: Learning from Sibling Harnesses for Generalizable Agent Training(Harness-S:向兄弟执行框架学习,实现可泛化的智能体训练) [01:24] 🎲 Random Is Almost Enough? Stable and Low-Cost Benchmark Compression via BestRand(随机抽样几乎就够了吗?通过 BestRand 实现稳定、低成本的基准压缩) [01:53] 🎨 TailSteer: Trajectory-Tail Steering for Drift-Resistant Multi-turn Image Editing(TailSteer:通过轨迹尾段引导实现抗漂移的多轮图像编辑) [02:24] 🖼️ WeLM-Image: A Unified Image Generation Model towards Visual Intelligence(WeLM-Image:迈向视觉智能的统一图像生成模型) [02:59] 🤖 SLA: Scaling Spatial-Language-Action Model via Continual Pre-training(SLA:通过持续预训练扩展空间—语言—动作模型) [03:34] 🧠 EvoMemState: Evolving Memory State for Personal Agents(EvoMemState:为个人智能体演化记忆状态) [04:08] 🔎 VisSearch-Bench: Evaluating Long Horizon Visual Search in Multimodal Agents(VisSearch-Bench:评估多模态智能体的长程视觉搜索) [04:38] ⚖️ Towards Foundation Decision Models: Drex(迈向基础决策模型:Drex) [05:08] 🌉 On the Opportunities and Challenges of Prefill-to-Autoregressive Cross-Model Transfer(论预填充到自回归的跨模型迁移:机遇与挑战) [05:37] 👤 EDGAR: Efficient Diffusion-Driven Gaussian Avatar Refinement via Synthetic Failure(EDGAR:通过合成失败实现高效的扩散驱动高斯头像细化) [06:05] 📝 HeLM: Hierarchical Explicit Language Memory for Non-Markovian Manipulation(HeLM:用于非马尔可夫操作的分层显式语言记忆) [06:37] 🎬 Future Representation Distillation for Imagination-free World Action Models(用于免想象世界动作模型的未来表征蒸馏) [07:11] 🤝 From Autonomous Agents to Cognitive Symbionts: Toward AI-Augmented Human Intelligence(从自主智能体到认知共生体:迈向 AI 增强的人类智能) [07:42] 🎧 Decode Before You Hear: Speculative Listening for Chunk-Interleaved Full-Duplex Speech Language Models(听到之前先解码:面向分块交错全双工语音语言模型的推测式聆听) [播客论文目录配图] 【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

8 分钟 44 次播放 1周前

【月末特辑】10月最火AI论文 | 把经验变成预见;让音画同步生成

【月末特辑】10月最火AI论文 | 把经验变成预见;让音画同步生成

HuggingFace 每日AI论文速递

⏱️ Highlights [00:28] 🔮 Self-Retrospection Distillation: Turning Post-hoc Experiences into Prior Foresight(自我回顾蒸馏:将事后经验转化为事前预见) [01:38] 🎬 Kandinsky 6.0 Video: Foundation Models for Synchronized Video and Audio Generation(Kandinsky 6.0 Video:用于同步视频与音频生成的基础模型) [02:49] 🌳 The Trinity:Capability-Driven Recursive Self-Improvement for Multimodal Agents(The Trinity:多模态智能体的能力驱动递归自我改进) [03:59] 🦾 PDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations(PDRVLA:具有物理动力学表征的视觉—语言—动作模型) [05:09] 🧭 The Teacher Is a Direction, Not a Destination: Extrapolating RL-Induced Representation Residuals in On-Policy Distillation(教师是方向,而非终点:在同策略蒸馏中外推强化学习诱导的表征残差) [06:21] 🔍 False Frontiers: Diagnosing and Mitigating Co-Cheating in Self-Evolving Search Agents(虚假前沿:诊断与缓解自我进化搜索智能体中的共同作弊) [07:32] ⚖️ Jeva: System One Models from First Principles(Jeva:从第一性原理构建系统一模型) [08:43] 🧠 MemT-1.0: Towards Memory-Native Evolution(MemT-1.0:迈向记忆原生进化) [09:53] 🤖 Has GPT-6-Astra Achieved Embodied Intelligence?(GPT-6-Astra已经实现具身智能了吗?) [11:03] 🎨 UniEvo-VL: An On-policy Self-Distillation Training Recipe for Multimodal Model Self-improvement(UniEvo-VL:用于多模态模型自我改进的同策略自蒸馏训练配方) [播客论文目录配图] 【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

12 分钟 46 次播放 1周前

播放列表

自动播放下一个

播放列表还是空的

去找些喜欢的单集添加进来吧