Blog
告别冗长 PDF，一文读懂最新顶会与核心期刊的创新价值。
Marginals Before Conditionals
Splat2Real: Novel-view Scaling for Physical AI with 3D Gaussian Splatting
Lost in Backpropagation: The LM Head is a Gradient Bottleneck
Spatial-TTT: Streaming Visual-based Spatial Intelligence with Test-Time Training
$Ψ_0$: An Open Foundation Model Towards Universal Humanoid Loco-Manipulation
IndexCache: Accelerating Sparse Attention via Cross-Layer Index Reuse
Neural Thickets: Diverse Task Experts Are Dense Around Pretrained Weights
RoboClaw: An Agentic Framework for Scalable Long-Horizon Robotic Tasks
ShotVerse: Advancing Cinematic Camera Control for Text-Driven Multi-Shot Video Creation
Trust Your Critic: Robust Reward Modeling and Reinforcement Learning for Faithful Image Editing and Generation
EVATok: Adaptive Length Video Tokenization for Efficient Visual Autoregressive Generation
XSkill: Continual Learning from Experience and Skills in Multimodal Agents
GGPT: Geometry Grounded Point Transformer
DVD: Deterministic Video Depth Estimation with Generative Priors
OmniStream: Mastering Perception, Reconstruction and Action in Continuous Streams
Linking Perception, Confidence and Accuracy in MLLMs
SaPaVe: Towards Active Perception and Manipulation in Vision-Language-Action Models for Robotics
DreamVideo-Omni: Omni-Motion Controlled Multi-Subject Video Customization with Latent Identity Reinforcement Learning
Examining Reasoning LLMs-as-Judges in Non-Verifiable LLM Post-Training
Attention Residuals
V-JEPA 2.1: Unlocking Dense Features in Video Self-Supervised Learning
Mamba-3: Improved Sequence Modeling using State Space Principles
Mixture-of-Depths Attention
Why AI systems don't learn and what to do about it: Lessons on autonomous learning from cognitive science
MetaClaw: Just Talk – An Agent That Meta-Learns and Evolves in the Wild
Online Experiential Learning for Language Models
Post-Training Local LLM Agents for Linux Privilege Escalation with Verifiable Rewards
AI Can Learn Scientific Taste
Complementary Reinforcement Learning
Demystifing Video Reasoning