Blog
No more endless PDFs. Discover the core value of the latest top-tier research in one article.
A Novel Unified Conditional Score-based Generative Framework for Multi-modal Medical Image Completion
Using vision language models for safety hazard identification in construction
Unmasking On-Policy Distillation: Where It Helps, Where It Hurts, and Why
UniPool: A Globally Shared Expert Pool for Mixture-of-Experts
Compute Optimal Tokenization
Shape anisotropy governs organization of active rods: Swarming, turbulence, flocking, and jamming.
Muon is Not That Special: Random or Inverted Spectra Work Just as Well
Efficient Pre-Training with Token Superposition
Constrained Stochastic Spectral Preconditioning Converges for Nonconvex Objectives
Attention Drift: What Autoregressive Speculative Decoding Models Learn *
Medical Diffusion: Denoising Diffusion Probabilistic Models for 3D Medical Image Generation
Shape anisotropy governs organization of active rods: Swarming, turbulence, flocking, and jamming.
Solve the Loop: Attractor Models for Language and Reasoning
Rethinking On-Policy Distillation
All Circuits Lead to Rome: Rethinking Functional Anisotropy in Circuit and Sheaf Discovery for LLMs
Deep Learning as Neural Low-Degree Filtering: A Spectral Theory of Hierarchical Feature Learning
High-fidelity collisional quantum gates with fermionic atoms
VGGT- Ω
Penegakan Hukum Korupsi di Indonesia: Perspektif Systematic Literature Review (2010-2023)
OpenDeepThink: Parallel Reasoning via Bradley–Terry Aggregation
SlimQwen: Exploring the Pruning and Distillation in Large MoE Model Pre-training
On Training in Imagination
Learning, Fast and Slow: Towards LLMs That Adapt Continually
SilverSpeak: Evading AI-Generated Text Detectors using Homoglyphs
Advancing LLM Reasoning Generalists with Preference Trees
Forecasting frontier language model agent capabilities
1003_Chasing Offensive Conduct in Social Networks A Reputation-Based Practical Approach for Frisber.
1006_Just Cast the Net, and Hopefully the Right Fish Swim into It Audience Management on Social Network Sites.
1011_Facebook ride connect.
1018_Audio-driven facial animation by joint end-to-end learning of pose and emotion.