Blog
No more endless PDFs. Discover the core value of the latest top-tier research in one article.
Ring-Zero: Scaling Zero RL to a Trillion Parameters for Emergent Reasoning
Video Generation Models are General-Purpose Vision Learners
4
Harness Handbook: Making Evolving Agent Harnesses Readable,Navigable, and Editable
Proxy Exploration and Reusable Guidance: A Modular LLM Post-Training Paradigm via Proxy-Guided Update Signals
Rethinking the Evaluation of Harness Evolution for Agents
Invariant Learning Dynamics of Transformers in Inductive Reasoning Tasks
Self-Improvements in Modern Agentic Systems: A Survey
Concurrent Image Understanding and Generation: Self-Correcting Coupled Markov Jump Processes
From Observation to Insight: Mechanistic World Models and the Quest for Autonomous Discovery
TRACE: Turn-level Reward Assignment via Credit Estimation for Long-Horizon Agents TRACE: TURN-LEVEL REWARD ASSIGNMENT VIA CREDIT ESTIMATION FOR LONG-HORIZON AGENTS
Flow Matching in Feature Space for Stochastic World Modeling
UniVR: Thinking in Visual Space for Unified Visual Reasoning
Can a Language Model Learn Facts Continually in Its Weights?
Understanding Reasoning from Pretraining to Post-Training
Verbalizable Representations Form a Global Workspace in Language Models
Frontier Language Models Struggle to Copy: Text Can Be Better Viewed in 2D
Hail: Mechanisms, monitoring, forecasting, damages, financial compensation systems, and prevention
From Memory to Skills: Evidence-Grounded Co-Evolution Governance for Long-Horizon LLM Agents
Enhancing Rubric-based RL via Self-Distillation
Self-Modifying Lean Proof Agents with Verifier-Grounded Benchmark Coevolution
Experience Memory Graph: One-Shot Error Correction for Agents
Scaling Laws for Hypernetwork-Based Knowledge Injection in Large Language Models
K12-KGraph: A Curriculum-Aligned Knowledge Graph for Benchmarking and Training Educational LLMs
PRO-LONG: Programmatic Memory Enables Long-Horizon Reasoning
CoLT: Teaching Multi-Modal Models to Think with Chain of Latent Thoughts
Hilbert Operator for Progressive Encoding (HOPE): A Mathematical Framework for Deconstructing Learned Representations in Deep Networks
NVIDIA-labs OO Agents: Native Python Object-Oriented Agents
OpenForgeRL: Train Harness-native Agents in Any Environment
JAXBench: Benchmarking Autonomous TPU Kernel Optimization