ARC Is a Vision Problem!

Solving a Million-Step LLM Task with Zero Errors

DataRater: Meta-Learned Dataset Curation

Mathematical exploration and discovery at scale

Kosmos: An AI Scientist for Autonomous Discovery

World Simulation with Video Foundation Models for Physical AI

Towards Robust Mathematical Reasoning

ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models

Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models

ImpossibleBench: Measuring LLMs’ Propensity of Exploiting Test Cases

Scaling Instruction-Based Video Editing with a High-Quality Synthetic Dataset

Reasoning with Sampling: Your Base Model is Smarter Than You Think

DeepSeek-OCR: Contexts Optical Compression

The Markovian Thinker

DeepDive: Advancing Deep Search Agents with Knowledge Graphs and Multi-Turn RL

Towards a Physics Foundation Model

Scalable Option Learning in High-Throughput Environments

Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning

Reverse-Engineered Reasoning for Open-Ended Generation

Scaling Performance of Large Language Model Pretraining