三模态从零训练，Agent RL稳定性破局

重点关注

• SkyReels-V4: Multi-modal Video-Audio Generation, Inpainting and Editing model score 7
  入选 HF Daily Papers；HF 热度: 46 upvotes (+4)
• ARLArena: A Unified Framework for Stable Agentic Reinforcement Learning score 10
  入选 HF Daily Papers；HF 热度: 22 upvotes (+4)；有代码实现；关键词(1): agentic
• Solaris: Building a Multiplayer Video World Model in Minecraft score 9
  入选 HF Daily Papers；HF 热度: 22 upvotes (+4)；有代码实现
• The Design Space of Tri-Modal Masked Diffusion Models score 8
  机构: Apple；入选 HF Daily Papers；HF 热度: 3 upvotes (+1)；关键词(3): scaling, fine-tuning, text-to-image
• GUI-Libra: Training Native GUI Agents to Reason and Act with Action-aware Supervision and Partially Verifiable RL score 9
  入选 HF Daily Papers；HF 热度: 14 upvotes (+3)；有代码实现；关键词(5): scaling, post-training, reasoning, open-source, data curation
• VecGlypher: Unified Vector Glyph Generation with Language Models score 9
  入选 HF Daily Papers；HF 热度: 11 upvotes (+3)；有代码实现；关键词(1): post-training
• Accelerating Diffusion via Hybrid Data-Pipeline Parallelism Based on Conditional Guidance Scheduling score 9
  入选 HF Daily Papers；HF 热度: 12 upvotes (+3)；有代码实现；关键词(1): latency
• From Statics to Dynamics: Physics-Aware Image Editing with Latent Transition Priors score 8
  入选 HF Daily Papers；HF 热度: 9 upvotes (+2)；有代码实现；关键词(2): reasoning, open-source
• World Guidance: World Modeling in Condition Space for Action Generation score 6
  入选 HF Daily Papers；HF 热度: 8 upvotes (+2)；关键词(1): vision-language
• Revisiting Text Ranking in Deep Research score 7
  入选 HF Daily Papers；HF 热度: 4 upvotes (+1)；有代码实现；关键词(1): open-source

也值得关注

• Asymptotically Fast Clebsch-Gordan Tensor Products with Vector Spherical Harmonics score 3
  机构: MIT
• Easy3E: Feed-Forward 3D Asset Editing via Rectified Voxel Flow score 3
  顶会接收: CVPR
• AHAN: Asymmetric Hierarchical Attention Network for Identical Twin Face Verification score 3
  顶会接收: AAAI
• Fair Model-based Clustering score 4
  关键词(1): scaling；顶会接收: AAAI
• Extending Sequence Length is Not All You Need: Effective Integration of Multimodal Signals for Gene Expression Prediction score 3
  顶会接收: ICLR
• Global-Aware Edge Prioritization for Pose Graph Initialization score 4
  关键词(1): edge；顶会接收: CVPR
• Lumosaic: Hyperspectral Video via Active Illumination and Coded-Exposure Pixels score 4
  关键词(1): real-time；顶会接收: CVPR
• Vision Transformers Need More Than Registers score 3
  顶会接收: CVPR
• DLT-Corpus: A Large-Scale Text Collection for the Distributed Ledger Technology Domain score 6
  入选 HF Daily Papers；HF 热度: 2 upvotes (+1)；有代码实现
• Evaluating the Usage of African-American Vernacular English in Large Language Models score 3
  机构: Yale
• Easy to Learn, Yet Hard to Forget: Towards Robust Unlearning Under Bias score 3
  顶会接收: AAAI
• UNet-Based Keypoint Regression for 3D Cone Localization in Autonomous Racing score 3
  顶会接收: ICCV
• AeroDGS: Physically Consistent Dynamic Gaussian Splatting for Single-Sequence Aerial 4D Reconstruction score 3
  顶会接收: CVPR