本期的 15 篇论文如下:
[00:23] 🖼 Aligned Novel View Image and Geometry Synthesis via Cross-modal Attention Instillation(基于跨模态注意力提炼的对齐新视角图像与几何体合成)
[01:02] 🛡 Effective Red-Teaming of Policy-Adherent Agents(有效对抗策略依从型智能体)
[01:39] 🔄 The Diffusion Duality(扩散二元性)
[02:20] 🤖 LiveCodeBench Pro: How Do Olympiad Medalists Judge LLMs in Competitive Programming?(LiveCodeBench Pro:奥林匹克竞赛奖牌获得者如何评价大型语言模型在算法竞赛中的表现?)
[03:09] 🧠 pLSTM: parallelizable Linear Source Transition Mark networks(pLSTM:可并行化的线性源转移马尔可夫网络)
[03:50] 🖼 A High-Quality Dataset and Reliable Evaluation for Interleaved Image-Text Generation(高质量的图文交错生成数据集与可靠评估)
[04:36] 🧠 Beyond Homogeneous Attention: Memory-Efficient LLMs via Fourier-Approximated KV Cache(超越同质注意力:通过傅里叶近似KV缓存实现内存高效的LLM)
[05:16] 🤖 SkillBlender: Towards Versatile Humanoid Whole-Body Loco-Manipulation via Skill Blending(SkillBlender: 面向通用人形机器人全身Loco-操作的技能融合)
[06:00] 🧠 SwS: Self-aware Weakness-driven Problem Synthesis in Reinforcement Learning for LLM Reasoning(SwS:基于自感知弱点驱动的问题合成,用于提升大型语言模型在强化学习中的推理能力)
[06:42] 🛡 Detecting Harmful Memes with Decoupled Understanding and Guided CoT Reasoning(利用解耦理解和引导式CoT推理检测有害模因)
[07:17] 🎬 DeepVideo-R1: Video Reinforcement Fine-Tuning via Difficulty-aware Regressive GRPO(DeepVideo-R1:通过难度感知回归GRPO进行视频强化微调)
[07:59] ⚙ Configurable Preference Tuning with Rubric-Guided Synthetic Data(基于规则引导合成数据的可配置偏好调整)
[08:41] 👁 ViCrit: A Verifiable Reinforcement Learning Proxy Task for Visual Perception in VLMs(ViCrit:一种用于VLM中视觉感知的可验证强化学习代理任务)
[09:29] 🔄 A Self-Refining Framework for Enhancing ASR Using TTS-Synthesized Data(一种利用TTS合成数据增强ASR的自精炼框架)
[10:16] 🔍 Dense Retrievers Can Fail on Simple Queries: Revealing The Granularity Dilemma of Embeddings(稠密检索器在简单查询上可能失效:揭示嵌入的粒度困境)
【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