HuggingFace 每日AI论文速递 - 节目列表

本期的 14 篇论文如下：[00:26] 🤖 MLLM can see? Dynamic Correction Decoding for Hallucination Mitigation（多模态大语言模型能看见吗？动态校正解码以减轻幻觉）[01:07] 🛠 MTU-Bench: A Multi-granularity Tool-Use Benchmark for Large Language Models（MTU-Bench：大型语言模型的多粒度工具使用基准）[01:47] 📚 LLM$\times$MapReduce: Simplified Long-Sequence Processing using Large Language Models（LLM×MapReduce：利用大型语言模型简化长序列处理）[02:25] 🛡 SecCodePLT: A Unified Platform for Evaluating the Security of Code GenAI（SecCodePLT：评估代码生成AI安全性的统一平台）[03:01] 📹 LVD-2M: A Long-take Video Dataset with Temporally Dense Captions（LVD-2M：一个带有时间密集标注的长镜头视频数据集）[03:44] 🧠 What Matters in Transformers? Not All Attention is Needed（Transformer中什么最重要？并非所有注意力机制都必要）[04:18] 🌟 GS^3: Efficient Relighting with Triple Gaussian Splatting（GS^3：高效的三重高斯点云重光照）[04:51] 🤯 Your Mixture-of-Experts LLM Is Secretly an Embedding Model For Free（你的混合专家大型语言模型实际上是一个免费的嵌入模型）[05:31] 🌍 Efficiently Democratizing Medical LLMs for 50 Languages via a Mixture of Language Family Experts（通过语言家族专家混合模型高效实现50种语言的医疗大语言模型民主化）[06:08] 🚀 SimBa: Simplicity Bias for Scaling Up Parameters in Deep Reinforcement Learning（SimBa：深度强化学习中扩展参数的简单性偏置）[06:43] 📊 Efficient Diffusion Models: A Comprehensive Survey from Principles to Practices（高效扩散模型：从原理到实践的综合调查）[07:14] 🤖 Towards Synergistic, Generalized, and Efficient Dual-System for Robotic Manipulation（面向协同、广义和高效的双系统机器人操作）[07:58] 🔄 Empirical Study of Mutual Reinforcement Effect and Application in Few-shot Text Classification Tasks via Prompt（互增强效应的实证研究及其在少样本文本分类任务中的应用通过提示）[08:37] 🌍 Towards Natural Image Matting in the Wild via Real-Scenario Prior（面向自然图像抠图的现实场景先验）【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速递在小宇宙查看该单集文稿

本期的 15 篇论文如下：[00:24] 🌐 MMIE: Massive Multimodal Interleaved Comprehension Benchmark for Large Vision-Language Models（大规模多模态交错理解基准测试）[01:06] 🤖 LOKI: A Comprehensive Synthetic Data Detection Benchmark using Large Multimodal Models（LOKI：基于大型多模态模型的综合合成数据检测基准）[02:01] 🔍 Toward General Instruction-Following Alignment for Retrieval-Augmented Generation（面向检索增强生成的通用指令遵循对齐）[02:36] 📊 MEGA-Bench: Scaling Multimodal Evaluation to over 500 Real-World Tasks（MEGA-Bench：将多模态评估扩展到500多个真实世界任务）[03:12] 🎥 Animate-X: Universal Character Image Animation with Enhanced Motion Representation（Animate-X：增强运动表示的通用角色图像动画）[04:02] 📚 Omni-MATH: A Universal Olympiad Level Mathematic Benchmark For Large Language Models（全能数学：面向大型语言模型的奥林匹克级数学基准）[04:44] 📚 LiveXiv -- A Multi-Modal Live Benchmark Based on Arxiv Papers Content（LiveXiv -- 基于Arxiv论文内容的多模态实时基准）[05:29] 🎥 Cavia: Camera-controllable Multi-view Video Diffusion with View-Integrated Attention（Cavia：具有视角控制的多视角视频扩散与视角集成注意力）[06:09] ⏳ TemporalBench: Benchmarking Fine-grained Temporal Understanding for Multimodal Video Models（时间轴基准：多模态视频模型细粒度时间理解评测）[06:58] 🌊 Semantic Image Inversion and Editing using Rectified Stochastic Differential Equations（基于校正随机微分方程的语义图像反演与编辑）[07:40] 📊 Rethinking Data Selection at Scale: Random Selection is Almost All You Need（重新思考大规模数据选择：随机选择几乎是你所需要的）[08:26] 🌲 Tree of Problems: Improving structured problem solving with compositionality（问题树：通过组合性改进结构化问题解决）[09:13] 📺 TVBench: Redesigning Video-Language Evaluation（TVBench：重塑视频语言评估）[09:54] 🤖 Generalizable Humanoid Manipulation with Improved 3D Diffusion Policies（可泛化的人形机器人操作：改进的三维扩散策略）[10:29] 📚 LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory（长时记忆评估：在长期交互记忆中评估聊天助手）【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速递在小宇宙查看该单集文稿

本期的 16 篇论文如下：[00:25] 🌐 Baichuan-Omni Technical Report（百川-Omni 技术报告）[00:59] 🖼 Meissonic: Revitalizing Masked Generative Transformers for Efficient High-Resolution Text-to-Image Synthesis（Meissonic：高效高分辨率文本到图像生成的掩码生成Transformer复兴）[01:41] 🔧 From Generalist to Specialist: Adapting Vision Language Models via Task-Specific Visual Instruction Tuning（从通才到专家：通过任务特定视觉指令调整适应视觉语言模型）[02:17] 🎨 EvolveDirector: Approaching Advanced Text-to-Image Generation with Large Vision-Language Models（进化导演：利用大规模视觉语言模型接近高级文本到图像生成）[02:53] 🧠 StructRAG: Boosting Knowledge Intensive Reasoning of LLMs via Inference-time Hybrid Information Structurization（结构化RAG：通过推理时混合信息结构化提升LLMs的知识密集型推理能力）[03:34] 📏 PositionID: LLMs can Control Lengths, Copy and Paste with Explicit Positional Awareness（大语言模型：具备显式位置感知的长度控制与复制粘贴）[04:11] 🌐 Semantic Score Distillation Sampling for Compositional Text-to-3D Generation（语义分数蒸馏采样用于组合式文本到3D生成）[04:47] 🧠 SuperCorrect: Supervising and Correcting Language Models with Error-Driven Insights（超级纠正：利用错误驱动的洞察力监督和纠正语言模型）[05:29] 🔄 Mechanistic Permutability: Match Features Across Layers（机制可置换性：跨层匹配特征）[06:07] 🤖 Multi-Agent Collaborative Data Selection for Efficient LLM Pretraining（多智能体协作数据选择以提高LLM预训练效率）[06:45] ⚡ KV Prediction for Improved Time to First Token（KV预测提升首次输出时间）[07:30] 🌐 ZeroComp: Zero-shot Object Compositing from Image Intrinsics via Diffusion（零样本对象合成：基于扩散的图像内在特性）[08:13] 🚨 MiRAGeNews: Multimodal Realistic AI-Generated News Detection（多模态现实AI生成新闻检测）[08:52] 🤖 DA-Code: Agent Data Science Code Generation Benchmark for Large Language Models（DA-Code：面向大型语言模型的代理数据科学代码生成基准）[09:30] 📈 I-Max: Maximize the Resolution Potential of Pre-trained Rectified Flow Transformers with Projected Flow（I-Max：最大化预训练校正流变换器的分辨率潜力与投影流）[10:12] 🧠 Mentor-KD: Making Small Language Models Better Multi-step Reasoners（导师-KD：使小型语言模型成为更好的多步推理者）【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速递在小宇宙查看该单集文稿

本期的 5 篇论文如下：[00:37] TOP1(🔥128) | 🔍 Differential Transformer（差分Transformer）[02:38] TOP2(🔥125) | ⚡ Addition is All You Need for Energy-efficient Language Models（加法即所需：高效能语言模型）[04:13] TOP3(🔥84) | 🌐 Aria: An Open Multimodal Native Mixture-of-Experts Model（Aria：一个开放的多模态原生混合专家模型）[06:18] TOP4(🔥73) | 🤖 GLEE: A Unified Framework and Benchmark for Language-based Economic Environments（GLEE：基于语言的经济环境统一框架与基准）[08:25] TOP5(🔥63) | 👤 Personalized Visual Instruction Tuning（个性化视觉指令微调）【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速递在小宇宙查看该单集文稿

本期的 21 篇论文如下：[00:25] 🧮 MathCoder2: Better Math Reasoning from Continued Pretraining on Model-translated Mathematical Code（MathCoder2：通过模型翻译的数学代码进行持续预训练以提升数学推理能力）[01:09] 🚀 PrefixQuant: Static Quantization Beats Dynamic through Prefixed Outliers in LLMs（前缀量化：静态量化通过LLMs中的前缀异常值超越动态量化）[01:59] 🤖 MLLM as Retriever: Interactively Learning Multimodal Retrieval for Embodied Agents（MLLM作为检索器：交互式学习多模态检索以增强具身代理）[02:33] 🎨 DICE: Discrete Inversion Enabling Controllable Editing for Multinomial Diffusion and Masked Generative Models（DICE：离散逆向可控编辑的多项扩散与掩码生成模型）[03:03] 🔄 Benchmarking Agentic Workflow Generation（代理工作流生成基准测试）[03:44] 🤖 Agent S: An Open Agentic Framework that Uses Computers Like a Human（Agent S：一个使用计算机如人类的开放代理框架）[04:23] 🔄 Rectified Diffusion: Straightness Is Not Your Need in Rectified Flow（修正扩散：在修正流中直线性并非必需）[04:55] 🤖 Intriguing Properties of Large Language and Vision Models（大型语言与视觉模型的引人特性）[05:35] 🎥 Progressive Autoregressive Video Diffusion Models（渐进式自回归视频扩散模型）[06:26] 🌲 Towards Self-Improvement of LLMs via MCTS: Leveraging Stepwise Knowledge with Curriculum Preference Learning（基于MCTS的LLMs自我改进：利用逐步知识与课程偏好学习）[07:10] 🌐 Preserving Multi-Modal Capabilities of Pre-trained VLMs for Improving Vision-Linguistic Compositionality（保留预训练视觉语言模型的多模态能力以提升视觉语言组合性）[07:50] 🤖 GLOV: Guided Large Language Models as Implicit Optimizers for Vision Language Models（GLOV：引导大型语言模型作为视觉语言模型的隐式优化器）[08:36] 🧩 SFTMix: Elevating Language Model Instruction Tuning with Mixup Recipe（SFTMix：利用Mixup方法提升语言模型指令微调）[09:15] 🔄 Emergent properties with repeated examples（重复示例的涌现特性）[09:57] 🤖 Optima: Optimizing Effectiveness and Efficiency for LLM-Based Multi-Agent System（优化基于LLM的多智能体系统的有效性与效率）[10:40] 🎲 Cheating Automatic LLM Benchmarks: Null Models Achieve High Win Rates（欺骗自动LLM基准测试：空模型实现高胜率）[11:14] 🌐 Everything Everywhere All at Once: LLMs can In-Context Learn Multiple Tasks in Superposition（无处不在同时进行：LLMs 可以在叠加状态下进行多任务上下文学习）[11:58] 🧬 LPZero: Language Model Zero-cost Proxy Search from Zero（LPZero：从零开始的零成本代理搜索）[12:41] 🌐 MotionGS: Exploring Explicit Motion Guidance for Deformable 3D Gaussian Splatting（MotionGS：探索显式运动引导的可变形3D高斯喷射）[13:15] 🔍 Scaling Up Your Kernels: Large Kernel Design in ConvNets towards Universal Representations（扩展你的卷积核：大卷积核设计在卷积神经网络中的通用表示）[13:51] 🖼 DART: Denoising Autoregressive Transformer for Scalable Text-to-Image Generation（DART：去噪自回归Transformer用于可扩展的文本到图像生成）【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速递在小宇宙查看该单集文稿

本期的 10 篇论文如下：[00:40] TOP1(🔥129) | 🤖 Training Language Models to Self-Correct via Reinforcement Learning（通过强化学习训练语言模型进行自我修正）[02:41] TOP2(🔥121) | 🚀 Qwen2.5-Coder Technical Report（Qwen2.5-Coder技术报告）[04:44] TOP3(🔥96) | 🌐 Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Multimodal Models（Molmo 和 PixMo：用于最先进多模态模型的开放权重和开放数据）[06:30] TOP4(🔥95) | 🖼 Guide-and-Rescale: Self-Guidance Mechanism for Effective Tuning-Free Real Image Editing（引导与重缩放：无调参自引导机制实现高效真实图像编辑）[08:23] TOP5(🔥86) | 🧠 Attention Heads of Large Language Models: A Survey（大型语言模型注意力头：一项综述）[10:17] TOP6(🔥85) | 🎥 Loopy: Taming Audio-Driven Portrait Avatar with Long-Term Motion Dependency（Loopy：驯服音频驱动的人像化身与长期运动依赖）[11:56] TOP7(🔥81) | 🌐 OmniGen: Unified Image Generation（全能生成：统一图像生成模型）[13:51] TOP8(🔥81) | 🧠 Emu3: Next-Token Prediction is All You Need（Emu3：下一个词预测是所有你需要的）[15:45] TOP9(🔥78) | 📄 General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model（通用OCR理论：通过统一端到端模型迈向OCR-2.0）[17:59] TOP10(🔥77) | 🧠 OLMoE: Open Mixture-of-Experts Language Models（OLMoE：开放式混合专家语言模型）【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速递在小宇宙查看该单集文稿

本期的 43 篇论文如下：[00:23] 🤖 GLEE: A Unified Framework and Benchmark for Language-based Economic Environments（GLEE：基于语言的经济环境统一框架与基准）[01:09] 👤 Personalized Visual Instruction Tuning（个性化视觉指令微调）[01:48] 🌍 Towards World Simulator: Crafting Physical Commonsense-Based Benchmark for Video Generation（迈向世界模拟器：基于物理常识的视频生成基准）[02:35] 🖼 IterComp: Iterative Composition-Aware Feedback Learning from Model Gallery for Text-to-Image Generation（迭代组合感知反馈学习：从模型库中提升文本到图像生成）[03:17] 🔍 Deciphering Cross-Modal Alignment in Large Vision-Language Models with Modality Integration Rate（解码大型视觉语言模型中的跨模态对齐与模态集成率）[03:54] 🌐 Aria: An Open Multimodal Native Mixture-of-Experts Model（Aria：一个开放的多模态原生混合专家模型）[04:29] 🌐 Pixtral 12B（Pixtral 12B）[05:09] 🎥 Pyramidal Flow Matching for Efficient Video Generative Modeling（金字塔流匹配用于高效视频生成建模）[05:49] 🔗 Unveiling the Backbone-Optimizer Coupling Bias in Visual Representation Learning（揭示视觉表示学习中的骨干-优化器耦合偏差）[06:29] 🎥 MM-Ego: Towards Building Egocentric Multimodal LLMs（MM-Ego：构建以自我为中心的多模态大型语言模型）[07:07] 🔄 One Initialization to Rule them All: Fine-tuning via Explained Variance Adaptation（一种初始化方法统治所有：通过解释方差适应进行微调）[07:51] 📖 Story-Adapter: A Training-free Iterative Framework for Long Story Visualization（故事适配器：一种无需训练的迭代框架用于长故事可视化）[08:33] 🚀 Self-Boosting Large Language Models with Synthetic Preference Data（利用合成偏好数据自我提升大型语言模型）[09:13] 🚀 Falcon Mamba: The First Competitive Attention-free 7B Language Model（猎鹰曼巴：首个无注意力机制的7B语言模型）[09:53] 🎨 TweedieMix: Improving Multi-Concept Fusion for Diffusion-based Image/Video Generation（TweedieMix：改进基于扩散的图像/视频生成中的多概念融合）[10:24] ⏳ Temporal Reasoning Transfer from Text to Video（从文本到视频的时间推理迁移）[10:54] 🎥 TRACE: Temporal Grounding Video LLM via Causal Event Modeling（TRACE：通过因果事件建模实现视频时间定位的大型语言模型）[11:30] 📊 Data Selection via Optimal Control for Language Models（通过最优控制进行语言模型数据选择）[12:07] 🤖 Response Tuning: Aligning Large Language Models without Instruction（响应调优：无需指令对齐大型语言模型）[12:49] 🤖 CursorCore: Assist Programming through Aligning Anything（CursorCore：通过对齐任何内容辅助编程）[13:36] 🎥 ViBiDSampler: Enhancing Video Interpolation Using Bidirectional Diffusion Sampler（ViBiDSampler：利用双向扩散采样器增强视频插值）[14:16] 🗣 Mixed-Session Conversation with Egocentric Memory（带有自我中心记忆的混合会话）[14:57] 🎮 ING-VP: MLLMs cannot Play Easy Vision-based Games Yet（ING-VP：多模态大语言模型在视觉游戏中的表现仍不尽人意）[15:41] 🔓 AutoDAN-Turbo: A Lifelong Agent for Strategy Self-Exploration to Jailbreak LLMs（AutoDAN-Turbo：一种用于策略自我探索以破解LLMs的终身代理）[16:26] 🎥 T2V-Turbo-v2: Enhancing Video Generation Model Post-Training through Data, Reward, and Conditional Guidance Design（T2V-Turbo-v2：通过数据、奖励和条件引导设计增强视频生成模型后训练）[17:00] 📖 Collective Critics for Creative Story Generation（创意故事生成的集体批评框架）[17:36] 🎵 Diversity-Rewarded CFG Distillation（多样性奖励的CFG蒸馏）[18:16] 🧠 Retrieval-Augmented Decision Transformer: External Memory for In-context RL（检索增强决策变压器：上下文强化学习的外部记忆）[18:57] 🎙 F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching（F5-TTS：基于流匹配生成流畅且忠实语音的童话生成器）[19:32] 🎹 FürElise: Capturing and Physically Synthesizing Hand Motions of Piano Performance（《致爱丽丝：捕捉并物理合成钢琴演奏手部动作》）[20:20] 🧠 Holistic Unlearning Benchmark: A Multi-Faceted Evaluation for Text-to-Image Diffusion Model Unlearning（整体遗忘基准：文本到图像扩散模型遗忘的多方面评估）[21:01] 🧬 Multimodal Large Language Models for Inverse Molecular Design with Retrosynthetic Planning（多模态大语言模型用于逆向分子设计与逆合成规划）[21:38] 🎥 BroadWay: Boost Your Text-to-Video Generation Model in a Training-free Way（BroadWay：无需训练提升文本到视频生成模型）[22:21] 🚨 Multimodal Situational Safety（多模态情境安全）[22:56] 💥 Hallucinating AI Hijacking Attack: Large Language Models and Malicious Code Recommenders（幻觉AI劫持攻击：大型语言模型与恶意代码推荐器）[23:38] 🛠 Seeker: Enhancing Exception Handling in Code with LLM-based Multi-Agent Approach（Seeker：利用基于LLM的多代理方法增强代码中的异常处理）[24:18] 🌐 Jointly Generating Multi-view Consistent PBR Textures using Collaborative Control（联合生成多视角一致的PBR纹理：协作控制方法）[24:55] 🤖 TinyEmo: Scaling down Emotional Reasoning via Metric Projection（TinyEmo：通过度量投影缩小情感推理）[25:29] 🧠 MentalArena: Self-play Training of Language Models for Diagnosis and Treatment of Mental Health Disorders（心理竞技场：通过自我对弈训练语言模型用于心理健康障碍的诊断与治疗）[26:08] 🎭 TextToon: Real-Time Text Toonify Head Avatar from Single Video（文本转卡通：从单视频实时生成卡通化头部虚拟形象）[26:49] 🤖 Do great minds think alike? Investigating Human-AI Complementarity in Question Answering with CAIMIRA（伟大的思想是否一致？探究CAIMIRA框架下的人机问答互补性）[27:28] 📊 MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering（MLE-bench：评估机器学习代理在机器学习工程中的表现）[28:03] 🧠 Does Spatial Cognition Emerge in Frontier Models?（空间认知在前沿模型中是否出现？）【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速递在小宇宙查看该单集文稿

本期的 9 篇论文如下：[00:28] 📚 LongGenBench: Long-context Generation Benchmark（长上下文生成基准：LongGenBench）[01:11] 🌐 $\textbf{Only-IF}$:Revealing the Decisive Effect of Instruction Diversity on Generalization（仅限IF：揭示指令多样性对泛化的决定性影响）[01:50] 📊 RevisEval: Improving LLM-as-a-Judge via Response-Adapted References（RevisEval：通过响应自适应参考改进LLM作为评判者）[02:35] 🌟 A Spark of Vision-Language Intelligence: 2-Dimensional Autoregressive Transformer for Efficient Finegrained Image Generation（视觉语言智能的火花：用于高效细粒度图像生成的二维自回归Transformer）[03:25] 🎥 Grounded-VideoLLM: Sharpening Fine-grained Temporal Grounding in Video Large Language Models（基于视频的大型语言模型：细化视频中的细粒度时间定位）[04:00] 🎨 ControlAR: Controllable Image Generation with Autoregressive Models（ControlAR：可控图像生成的自回归模型）[04:45] 🔍 Hyper-multi-step: The Truth Behind Difficult Long-context Tasks（超多步：困难长上下文任务背后的真相）[05:21] 🤖 MA-RLHF: Reinforcement Learning from Human Feedback with Macro Actions（MA-RLHF：基于宏动作的人类反馈强化学习）[06:03] 📊 EBES: Easy Benchmarking for Event Sequences（EBES：事件序列的简易基准测试）【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速递在小宇宙查看该单集文稿

本期的 21 篇论文如下：[00:26] 🔍 Differential Transformer（差分Transformer）[01:04] 🧠 LLMs Know More Than They Show: On the Intrinsic Representation of LLM Hallucinations（大语言模型知多于表：关于LLM幻觉的内在表征）[01:50] 📹 VideoGuide: Improving Video Diffusion Models without Training Through a Teacher's Guide（视频指南：通过教师指导提升视频扩散模型无需训练）[02:28] 📈 FAN: Fourier Analysis Networks（傅里叶分析网络）[03:05] 🏥 Named Clinical Entity Recognition Benchmark（命名临床实体识别基准）[03:37] 🔬 ScienceAgentBench: Toward Rigorous Assessment of Language Agents for Data-Driven Scientific Discovery（科学智能基准：面向数据驱动科学发现的语言智能体严格评估）[04:19] 🎶 UniMuMo: Unified Text, Music and Motion Generation（统一文本、音乐与动作生成）[04:55] 🔍 TLDR: Token-Level Detective Reward Model for Large Vision Language Models（TLDR：大视觉语言模型的令牌级侦探奖励模型）[05:35] 🎵 Presto! Distilling Steps and Layers for Accelerating Music Generation（快速！加速音乐生成的步骤和层级蒸馏）[06:08] 🖥 Navigating the Digital World as Humans Do: Universal Visual Grounding for GUI Agents（像人类一样导航数字世界：GUI代理的通用视觉基础）[06:49] 🖼 OmniBooth: Learning Latent Control for Image Synthesis with Multi-modal Instruction（全能展台：通过多模态指令学习图像合成的潜在控制）[07:29] 🌀 MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motion（MonST3R：一种在动态场景中估计几何的简单方法）[08:09] 🧠 LLaMA-Berry: Pairwise Optimization for O1-like Olympiad-Level Mathematical Reasoning（LLaMA-Berry：O1类奥林匹克级数学推理的成对优化）[08:50] 📊 MathHay: An Automated Benchmark for Long-Context Mathematical Reasoning in LLMs（MathHay：LLMs长上下文数学推理自动化基准）[09:39] 📊 GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models（GSM-符号化：理解大型语言模型在数学推理中的局限性）[10:34] 🤖 Autonomous Character-Scene Interaction Synthesis from Text Instruction（从文本指令自主合成角色场景互动）[11:12] 🧩 TurtleBench: Evaluating Top Language Models via Real-World Yes/No Puzzles（TurtleBench：通过真实世界的Yes/No谜题评估顶级语言模型）[12:00] 🤖 Grounding Language in Multi-Perspective Referential Communication（多视角指称通信中的语言接地）[12:48] 🎯 SePPO: Semi-Policy Preference Optimization for Diffusion Alignment（SePPO：扩散模型对齐的半策略偏好优化）[13:25] 🧩 What Matters for Model Merging at Scale?（大规模模型合并的关键因素是什么？）[14:02] 📊 SELECT: A Large-Scale Benchmark of Data Curation Strategies for Image Classification（SELECT：图像分类数据策展策略的大规模基准）【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速递在小宇宙查看该单集文稿

本期的 12 篇论文如下：[00:25] ⚡ Addition is All You Need for Energy-efficient Language Models（加法即所需：高效能语言模型）[01:03] 🧠 NL-Eye: Abductive NLI for Images（NL-Eye：图像的溯因自然语言推理）[01:40] 🔍 Selective Attention Improves Transformer（选择性注意力提升Transformer）[02:17] ⚡ Accelerating Auto-regressive Text-to-Image Generation with Training-free Speculative Jacobi Decoding（加速自回归文本到图像生成：无训练的推测性雅可比解码）[02:48] 🤖 Tutor CoPilot: A Human-AI Approach for Scaling Real-Time Expertise（导师助手：一种用于扩展实时专家知识的人机协作方法）[03:27] 🩺 A Comprehensive Survey of Mamba Architectures for Medical Image Analysis: Classification, Segmentation, Restoration and Beyond（医学图像分析中的Mamba架构综合调查：分类、分割、恢复及超越）[04:12] 🎨 RoCoTex: A Robust Method for Consistent Texture Synthesis with Diffusion Models（RoCoTex：一种基于扩散模型的鲁棒一致纹理合成方法）[04:59] 🧠 Erasing Conceptual Knowledge from Language Models（从语言模型中消除概念知识）[05:37] 📈 MIGA: Mixture-of-Experts with Group Aggregation for Stock Market Prediction（MIGA：基于专家组聚合的混合模型用于股票市场预测）[06:16] 🤖 CANVAS: Commonsense-Aware Navigation System for Intuitive Human-Robot Interaction（CANVAS：常识感知导航系统用于直观人机交互）[06:54] 🌳 NRGBoost: Energy-Based Generative Boosted Trees（NRGBoost：基于能量的生成增强树）[07:37] 🤖 GenSim2: Scaling Robot Data Generation with Multi-modal and Reasoning LLMs（GenSim2：利用多模态和推理LLMs扩展机器人数据生成）【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速递在小宇宙查看该单集文稿

本期的 5 篇论文如下：[00:47] TOP1(🔥73) | 🧠 Emu3: Next-Token Prediction is All You Need（Emu3：下一个词预测是所有你需要的）[02:42] TOP2(🔥48) | 🔗 Law of the Weakest Link: Cross Capabilities of Large Language Models（最弱环节定律：大型语言模型的跨能力）[04:26] TOP3(🔥45) | 🌐 MIO: A Foundation Model on Multimodal Tokens（MIO：基于多模态标记的基础模型）[06:26] TOP4(🔥44) | 🌐 Revisit Large-Scale Image-Caption Data in Pre-training Multimodal Foundation Models（重访大规模图像-标题数据在预训练多模态基础模型中的应用）[08:27] TOP5(🔥43) | 🧠 MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning（MM1.5：多模态大语言模型微调的方法、分析与洞察）【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速递在小宇宙查看该单集文稿

本期的 19 篇论文如下：[00:24] 🔄 Revisit Large-Scale Image-Caption Data in Pre-training Multimodal Foundation Models（重新审视大规模图像-文本数据在多模态基础模型预训练中的作用）[01:04] 🎥 Loong: Generating Minute-level Long Videos with Autoregressive Language Models（使用自回归语言模型生成分钟级长视频）[01:39] 🎥 Video Instruction Tuning With Synthetic Data（使用合成数据进行视频指令调优）[02:18] 🧐 LLaVA-Critic: Learning to Evaluate Multimodal Models（LLaVA-Critic：学习评估多模态模型）[02:56] 🔍 Contrastive Localized Language-Image Pre-Training（对比本地化语言-图像预训练）[03:31] 🌱 VinePPO: Unlocking RL Potential For LLM Reasoning Through Refined Credit Assignment（VinePPO：通过精细化的信用分配解锁LLM推理的RL潜力）[04:07] 🌟 Depth Pro: Sharp Monocular Metric Depth in Less Than a Second（Depth Pro：不到一秒内实现锐利的单目度量深度）[04:51] 🔗 Large Language Models as Markov Chains（大型语言模型作为马尔可夫链）[05:26] 🧠 CLIP-MoE: Towards Building Mixture of Experts for CLIP with Diversified Multiplet Upcycling（CLIP-MoE：通过多样化多重升级构建CLIP的专家混合模型）[06:03] 🔄 Eliminating Oversaturation and Artifacts of High Guidance Scales in Diffusion Models（消除扩散模型中高指导尺度引起的过饱和和伪影）[06:51] 🔄 Training Language Models on Synthetic Edit Sequences Improves Code Synthesis（在合成编辑序列上训练语言模型改进代码合成）[07:36] ⚡ SageAttention: Accurate 8-Bit Attention for Plug-and-play Inference Acceleration（SageAttention：用于即插即用推理加速的精确8位注意力机制）[08:14] 🌐 MVGS: Multi-view-regulated Gaussian Splatting for Novel View Synthesis（MVGS：多视角调节的高斯喷射用于新视角合成）[08:54] 📚 L-CiteEval: Do Long-Context Models Truly Leverage Context for Responding?（L-CiteEval：长上下文模型是否真正利用上下文进行响应？）[09:38] 🩺 MedVisionLlama: Leveraging Pre-Trained Large Language Model Layers to Enhance Medical Image Segmentation（利用预训练大型语言模型层增强医学图像分割）[10:24] 🎥 Vinoground: Scrutinizing LMMs over Dense Temporal Reasoning with Short Videos（Vinoground: 通过短视频密集时间推理审视大型多模态模型）[11:01] 🗣 Distilling an End-to-End Voice Assistant Without Instruction Training Data（无需指令训练数据的端到端语音助手蒸馏）[11:46] ♟ Learning the Latent Rules of a Game from Data: A Chess Story（从数据中学习游戏的潜在规则：一个国际象棋的故事）[12:29] 🎵 Synthio: Augmenting Small-Scale Audio Classification Datasets with Synthetic Data（Synthio：使用合成数据增强小规模音频分类数据集）【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速递在小宇宙查看该单集文稿