Research9
AI Systems Nearing Self-Improvement Capabilities
AI is on the cusp of automating its own research and development, paving the way for recursive self-improvement.
Import AI (Jack Clark)May 18
Read more
Research
Research
20 articles in this category
May 18, 2026
Research
6
Import AI 457: AI stuxnet; cursed Muon optimizer; and positive alignment
This issue of Import AI provides updates on diverse AI research topics, including security vulnerabilities, optimization techniques, and alignment strategies.
Import AI (Jack Clark)May 18
Read more
Research
8
Hugging Face and IBM Launch Open Agent Leaderboard for AI Agents
The introduction of a public, standardized leaderboard for AI agents is critical for transparent evaluation and accelerating progress in agent development.
Hugging Face BlogMay 18
Read more
April 30, 2026
Research
8
Enabling a new model for healthcare with AI co-clinician
Google DeepMind is pioneering research into an "AI co-clinician" to augment human medical expertise and transform healthcare delivery.
Google DeepMindApr 30
Read more
April 25, 2026
Research
8
Anthropic Creates Test Marketplace for Agent-on-Agent Commerce
Anthropic's experiment demonstrates the potential for AI agents to autonomously engage in real-world commerce, negotiating and completing transactions for actual goods and money.
TechCrunch AIApr 25
Read more
April 24, 2026
Research
8
DeepSeek Previews New AI Model Closing Gap with Frontier Models
DeepSeek's new AI models claim significant performance and efficiency improvements, nearing the capabilities of top frontier models on reasoning benchmarks.
TechCrunch AIApr 24
Read more
Research
8
DeepSeek-V4: New LLM Family Features 1 Million Token Context Window
DeepSeek-V4 significantly advances LLM capabilities with its massive 1 million token context window and strong agentic performance, making it a powerful new tool for AI development.
Hugging Face BlogApr 24
Read more
April 23, 2026
Research
8
Decoupled DiLoCo: A new frontier for resilient, distributed AI training
Decoupled DiLoCo offers a resilient and decoupled methodology for distributed AI training, promising more robust and efficient large-scale machine learning.
Google DeepMindApr 23
Read more
April 22, 2026
April 21, 2026
Research
8
AI's Physical World Challenge: The Promise of World Models
World models are considered a fundamental step for AI to achieve robust understanding and interaction within the physical world, bridging the gap between digital mastery and real-world competence.
MIT Technology Review AIApr 21
Read more
Research
8
ChatGPT's New Images 2.0 Model Excels at Generating Text Within Images
OpenAI's Images 2.0 model for ChatGPT showcases a notable leap in AI image generation, particularly its surprising ability to accurately render text within generated images.
TechCrunch AIApr 21
Read more
April 20, 2026
Research7
Automating AI Alignment Research, Chinese Model Safety, & HiFloat4 Discussed
The issue highlights the ongoing scientific and ethical challenges in advanced AI development, juxtaposed with broader market speculation about AI's long-term societal impact.
Import AI (Jack Clark)Apr 20
Read more
April 17, 2026
Research7
Ecom-RLVE: Adaptive Verifiable Environments for E-Commerce AI Agents
Ecom-RLVE offers a new research-backed approach to building more reliable and adaptable conversational AI agents for e-commerce through verifiable learning environments.
Hugging Face BlogApr 17
Read more
Research7
NVIDIA Nemotron-OCR-V2: Fast Multilingual OCR Model Built with Synthetic Data
NVIDIA's Nemotron-OCR-V2 leverages synthetic data generation to create a highly effective and fast multilingual OCR model, overcoming limitations of real-world data scarcity.
Hugging Face BlogApr 17
Read more
April 16, 2026
Research
8
Physical Intelligence Unveils π0.7 Robot Brain Learning Untaught Tasks
Physical Intelligence's π0.7 model demonstrates a novel capability for robots to learn and execute tasks without explicit prior training, advancing the pursuit of general-purpose AI in robotics.
TechCrunch AIApr 16
Read more
Research
8
OpenAI Introduces GPT-Rosalind for Life Sciences Research
OpenAI's GPT-Rosalind is a new, specialized AI model aimed at significantly advancing research capabilities across various life science domains.
OpenAI BlogApr 16
Read more
April 15, 2026
Research
8
Inside VAKRA: Reasoning, Tool Use, and Failure Modes of Agents
The VAKRA benchmark offers crucial insights into the current limitations of AI agents in reasoning and tool use, providing a clear roadmap for future research and development efforts.
Hugging Face BlogApr 15
Read more
April 14, 2026
Research7
Gemini Robotics-ER 1.6 Enhances Embodied Reasoning for Real-World Tasks
Gemini Robotics-ER 1.6 significantly advances robot intelligence by improving spatial and multi-view understanding, enabling more effective real-world task performance.
Google DeepMindApr 14
Read more
April 13, 2026
Research9
AI Achieves One-Shot Learning, Mastering Tasks from Single Examples
The development of one-shot learning capabilities marks a crucial step in AI's ability to generalize and learn efficiently from minimal data, mirroring human cognitive processes.
The Batch (Andrew Ng)Apr 13
Read more
Research
8
Stanford's 2026 AI Index Released, Offering Key Industry Insights
The 2026 AI Index serves as an essential annual report for understanding the multifaceted and often contradictory developments in the field of artificial intelligence through data-driven analysis.
MIT Technology Review AIApr 13
Read more