3B参数奥赛三金，768维离散token生成可行

重点关注

• Cubic Discrete Diffusion: Discrete Visual Generation on High-Dimensional Representation Tokens score 13
  入选 HF Daily Papers；HF 热度: 28 upvotes (+4)；有代码实现；关键词(1): scaling；顶会接收: CVPR
• Nemotron-Cascade 2: Post-Training LLMs with Cascade RL and Multi-Domain On-Policy Distillation score 11
  机构: NVIDIA；入选 HF Daily Papers；HF 热度: 41 upvotes (+4)；关键词(6): distillation, post-training, MoE, agentic, coding
• Generation Models Know Space: Unleashing Implicit 3D Priors for Scene Understanding score 10
  入选 HF Daily Papers；HF 热度: 78 upvotes (+4)；有代码实现；关键词(2): reasoning, embodied
• SAMA: Factorized Semantic Anchoring and Motion Alignment for Instruction-Guided Video Editing score 10
  入选 HF Daily Papers；HF 热度: 59 upvotes (+4)；有代码实现；关键词(3): fine-tuning, pre-training, open-source
• 3DreamBooth: High-Fidelity 3D Subject-Driven Video Generation Model score 10
  入选 HF Daily Papers；HF 热度: 47 upvotes (+4)；有代码实现；关键词(3): production, fine-tune, pre-training
• FASTER: Rethinking Real-Time Flow VLAs score 10
  入选 HF Daily Papers；HF 热度: 45 upvotes (+4)；有代码实现；关键词(3): latency, real-time, vision-language
• MonoArt: Progressive Structural Reasoning for Monocular Articulated 3D Reconstruction score 10
  入选 HF Daily Papers；HF 热度: 28 upvotes (+4)；有代码实现；关键词(1): reasoning
• LVOmniBench: Pioneering Long Audio-Video Understanding Evaluation for Omnimodal LLMs score 10
  入选 HF Daily Papers；HF 热度: 27 upvotes (+4)；有代码实现；关键词(1): open-source
• Bridging Semantic and Kinematic Conditions with Diffusion-based Discrete Motion Tokenizer score 9
  入选 HF Daily Papers；HF 热度: 36 upvotes (+4)；有代码实现
• Memento-Skills: Let Agents Design Agents score 9
  入选 HF Daily Papers；HF 热度: 31 upvotes (+4)；有代码实现

也值得关注

• Cross-Domain Demo-to-Code via Neurosymbolic Counterfactual Reasoning score 4
  关键词(4): deployment, state space, reasoning, vision-language；顶会接收: CVPR
• Remedying Target-Domain Astigmatism for Cross-Domain Few-Shot Object Detection score 4
  关键词(1): fine-tuning；顶会接收: CVPR
• CausalVAD: De-confounding End-to-End Autonomous Driving via Causal Intervention score 4
  关键词(1): lightweight；顶会接收: CVPR
• ZEBRAARENA: A Diagnostic Simulation Environment for Studying Reasoning-Action Coupling in Tool-Augmented LLMs score 4
  机构: Carnegie Mellon；关键词(2): tool use, reasoning
• 6Bit-Diffusion: Inference-Time Mixed-Precision Quantization for Video Diffusion Models score 4
  机构: Tsinghua；关键词(5): lightweight, compression, quantization, deployment, post-training
• DA-Mamba: Learning Domain-Aware State Space Model for Global-Local Alignment in Domain Adaptive Object Detection score 4
  关键词(3): deployment, state space, mamba；顶会接收: CVPR
• dTRPO: Trajectory Reduction in Policy Optimization of Diffusion Large Language Models score 4
  机构: Meta；关键词(2): coding, reasoning
• Agentic Business Process Management: A Research Manifesto score 4
  机构: Oxford；关键词(1): agentic
• DROID-SLAM in the Wild score 4
  机构: ETH Zurich；关键词(1): real-time
• CustomTex: High-fidelity Indoor Scene Texturing via Multi-Reference Customization score 4
  关键词(1): distillation；顶会接收: CVPR
• Optimal Splitting of Language Models from Mixtures to Specialized Domains score 4
  机构: EPFL；关键词(3): scaling, pretraining, reasoning
• DreamPartGen: Semantically Grounded Part-Level 3D Generation via Collaborative Latent Denoising score 4
  入选 HF Daily Papers；关键词(1): reasoning
• Cell-Type Prototype-Informed Neural Network for Gene Expression Estimation from Pathology Images score 3
  顶会接收: CVPR
• CAPSUL: A Comprehensive Human Protein Benchmark for Subcellular Localization score 3
  顶会接收: ICLR
• Elastic Weight Consolidation Done Right for Continual Learning score 3
  顶会接收: CVPR