Blog
No more endless PDFs. Discover the core value of the latest top-tier research in one article.
There Will Be a Scientific Theory of Deep Learning
Intelligent Parameter Decision-Making and Multi-objective Prediction for Multi-layer and Multi-pass LDED Process
There Will Be a Scientific Theory of Deep Learning
Low-Rank Adaptation Redux for Large Models
The Linear Centroids Hypothesis: How Deep Network Features Represent Data
Image Generators are Generalist Vision Learners
Hyperloop Transformers
TIP: Token Importance in On-Policy Distillation
LLaDA2.0-Uni: Unifying Multimodal Understanding and Generation with Diffusion Large Language Model
Agent-World: Scaling Real-World Environment Synthesis for Evolving General Agent Intelligence
The Recurrent Transformer: Greater Effective Depth and Efficient Decoding
SWE-chat: Coding Agent Interactions From Real Users in the Wild
SGD at the Edge of Stability: The Stochastic Sharpness Gap
Thinking Without Words: Efficient Latent Reasoning with Abstract Chain-of-Thought
Evaluating creative ideas: Processes, standards, and context
There Will Be a Scientific Theory of Deep Learning
Agentic World Modeling: Foundations, Capabilities, Laws, and Beyond
SFT-then-RL Outperforms Mixed-Policy Methods for LLM Reasoning
Tuna-2: Pixel Embeddings Beat Vision Encoders for Multimodal Understanding and Generation
Universal Transformers Need Memory: Depth-State Trade-offs in Adaptive Recursive Reasoning
Spend Less, Fit Better: Budget-Efficient Scaling Law Fitting via Active Experiment Selection
There Will Be a Scientific Theory of Deep Learning
From Skills to Talent: Organising Heterogeneous Agents as a Real-World Company
High-fidelity collisional quantum gates with fermionic atoms
Machining of ZrO2 ceramics with PCD and CBN cutting tools
How Well Does Generative Recommendation Generalize?
Incompressible Knowledge Probes: Estimating Black-Box LLM Parameter Counts via Factual Capacity
AdaExplore: Failure-Driven Adaptation and Diversity-Preserving Search for Efficient Kernel Generation
When Errors Can Be Beneficial: A Categorization of Imperfect Rewards for Policy Gradient
Introspection Adapters: Training LLMs to Report Their Learned Behaviors