Blog
No more endless PDFs. Discover the core value of the latest top-tier research in one article.
Kinema4D: Kinematic 4D World Modeling for Spatiotemporal Embodied Simulation
Mirror Descent on Riemannian Manifolds
The Unreasonable Effectiveness of Text Embedding Interpolation for Continuous Image Steering
MolmoB0T: Large-Scale Simulation Enables Zero-Shot Manipulation
VectorWorld: Efficient Streaming World Model via Diffusion Flow on Vector Graphs
FineViT: Progressively Unlocking Fine-Grained Perception with Dense Recaptions
Internalizing Agency from Reflective Experience
Toward bootstrapping tensor-network contractions
RoboForge: Physically Optimized Text-guided Whole-Body Locomotion for Humanoids
AdaMem: Adaptive User-Centric Memory for Long-Horizon Dialogue Agents
Listening to the Echo: User-Reaction Aware Policy Optimization via Scalar-Verbal Hybrid Reinforcement Learning
DriveTok: 3D Driving Scene Tokenization for Unified Multi-View Reconstruction and Understanding
Video-CoE: Reinforcing Video Event Prediction via Chain of Events
PowerFlow: Unlocking the Dual Nature of LLMs via Principled Distribution Matching
MOSS-TTS Technical Report
What Shape is the Inflationary Bispectrum?
Simulation Distillation: Pretraining World Models in Simulation for Rapid Real-World Adaptation
Mi:dm K 2.5 Pro
PJB: A Reasoning-Aware Benchmark for Person-Job Retrieval
SegviGen: Repurposing 3D Generative Model for Part Segmentation
Quantum Advantage: a Tensor Network Perspective
SAGE: Multi-Agent Self-Evolution for LLM Reasoning
Temporal Gains, Spatial Costs: Revisiting Video Fine-Tuning in Multimodal Large Language Models
REAL: Regression-Aware Reinforcement Learning for LLM-as-a-Judge
Identity as Presence: Towards Appearance and Voice Personalized Joint Audio-Video Generation
The Silent Thought: Modeling Internal Cognition in Full-Duplex Spoken Dialogue Models via Latent Reasoning
Fast SAM 3D Body: Accelerating SAM 3D Body for Real-Time Full-Body Human Mesh Recovery
DROID-SLAM in the Wild
Photon-echo synchronization and quantum state transfer in short quantum links
3DreamBooth: High-Fidelity 3D Subject-Driven Video Generation Model