OpenAI Launches Visual Reasoning Benchmark for AI
OpenAI unveils a new visual reasoning benchmark designed to stress-test multimodal AI systems on complex perception task…
📂 Research
Latest articles in Research
TRI Uses Diffusion Policy for Dexterous Robots
Toyota Research Institute demonstrates diffusion policy models enabling robots to perform complex manipulation tasks wit…
Sony Research Unveils AI for Autonomous Game NPCs
Sony Research Tokyo introduces a multimodal AI system that autonomously creates and controls game characters with unprec…
0% Pass Rate: New Benchmark Stumps All AI Models
The creators of SWE-Bench release a devastating new benchmark where Claude, GPT, and Gemini all score zero, exposing fun…
DeepMind Taps Eve Online to Crack AI Long-Term Planning
Google DeepMind acquires minority stake in Fenris Creations, plans to use Eve Online's complex virtual world to train AI…
RLHF Evolves Into Constitutional AI Training
The AI alignment landscape shifts as Constitutional AI methods begin replacing traditional RLHF, promising scalable and …
Oxford Study: Transformer Alternatives Could Break Context Limits
New Oxford research explores architectures beyond Transformers that may solve the quadratic scaling problem limiting con…
NVIDIA Unveils Real-Time NeRF for AR Apps
NVIDIA Research introduces a breakthrough neural radiance field system capable of rendering photorealistic 3D scenes in …
Diffusion Transformer Unifies Image, Video, and 3D
A new Diffusion Transformer architecture promises to merge image, video, and 3D generation into a single unified model, …
Google Achieves Quantum-Enhanced ML on 100 Qubits
Google Research demonstrates quantum-enhanced machine learning on a 100-qubit processor, marking a major step toward pra…
MIT Achieves Breakthrough in Energy-Efficient AI Training
MIT researchers unveil a novel training method that cuts neural network energy consumption by up to 75%, potentially sav…
OpenAI Reveals Emergent Tool Use in AI Agents
OpenAI publishes new research showing AI agents spontaneously develop tool use and complex strategies in multi-agent sim…