2025.12.08 | 自对抗一步生成;外挂评审迭代编辑

2025.12.08 | 自对抗一步生成;外挂评审迭代编辑

10分钟 ·
播放数104
·
评论数0

本期的 15 篇论文如下:

00:19 ⚡ TwinFlow: Realizing One-step Generation on Large Models with Self-adversarial Flows(TwinFlow:基于自对抗流实现大模型的一步生成)

00:49 🤔 EditThinker: Unlocking Iterative Reasoning for Any Image Editor(EditThinker:为任意图像编辑器解锁迭代推理能力)

01:26 🎨 PaCo-RL: Advancing Reinforcement Learning for Consistent Image Generation with Pairwise Reward Modeling(PaCo-RL:通过成对奖励建模推进强化学习在一致性图像生成中的应用)

02:05 📈 From Imitation to Discrimination: Toward A Generalized Curriculum Advantage Mechanism Enhancing Cross-Domain Reasoning Tasks(从模仿到判别:一种增强跨领域推理任务的通用课程优势机制)

02:55 ⚖ Entropy Ratio Clipping as a Soft Global Constraint for Stable Reinforcement Learning(熵比率裁剪:一种用于稳定强化学习的软性全局约束)

03:38 🎬 Joint 3D Geometry Reconstruction and Motion Generation for 4D Synthesis from a Single Image(基于单张图像的联合三维几何重建与运动生成以实现四维合成)

04:15 🧠 COOPER: A Unified Model for Cooperative Perception and Reasoning in Spatial Intelligence(COOPER:空间智能中协同感知与推理的统一模型)

04:45 🎨 RealGen: Photorealistic Text-to-Image Generation via Detector-Guided Rewards(RealGen:通过检测器引导的奖励实现逼真的文本到图像生成)

05:16 🔍 ReVSeg: Incentivizing the Reasoning Chain for Video Segmentation with Reinforcement Learning(ReVSeg:利用强化学习激励视频分割中的推理链)

05:49 🎥 World Models That Know When They Don't Know: Controllable Video Generation with Calibrated Uncertainty(知晓自身不确定性的世界模型:具有校准不确定性的可控视频生成)

06:24 🎮 SpaceControl: Introducing Test-Time Spatial Control to 3D Generative Modeling(SpaceControl:为3D生成模型引入测试时空间控制)

07:14 🤖 Self-Improving VLM Judges Without Human Annotations(无需人工标注的自改进视觉语言模型评判器)

07:54 🎬 SCAIL: Towards Studio-Grade Character Animation via In-Context Learning of 3D-Consistent Pose Representations(SCAIL:通过三维一致姿态表征的上下文学习实现影视级角色动画)

08:30 🤝 AI & Human Co-Improvement for Safer Co-Superintelligence(人工智能与人类协同进化以实现更安全的协同超级智能)

09:08 🎬 ProPhy: Progressive Physical Alignment for Dynamic World Simulation(ProPhy:面向动态世界模拟的渐进式物理对齐框架)

【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