98 articles
In-depth coverage of new AI research — papers, benchmarks, and breakthroughs from leading labs and academia, summarized for fast reading and grounded in the methods that matter.
Adrian de Wynter, a Microsoft researcher, created a working neural network in Age of Empires II using goats to critique AI science, revealing the game's mechanics can simulate computational processes.
Nvidia's ENPIRE project enables robots to train themselves through AI coding agents, achieving up to 99% success on complex tasks like pin insertion and cable tie closing.
Google's AMIE AI system, tested in a blinded study, matched 21 doctors in managing chronic conditions and scored higher in plan preciseness and guideline alignment.
OpenAI used Deployment Simulation to improve estimates of undesired model behavior in GPT-5 series models, reducing risks of model detection during testing.
AWS unveiled P-EAGLE, a new method for parallel speculative decoding, which boosts throughput by up to 1.69x on real-world benchmarks. The technique is now available in Amazon SageMaker JumpStart.
AMD announced MLPerf Training v6.0 results on June 16, 2026, showcasing performance on MI325X, MI350X, and MI355X Instinct GPUs.
The Institute of the Estonian Language tested 60 AI models against 75 Russian propaganda questions, finding Anthropic's Claude models top the list with scores over 95.
A new benchmark shows AI coding agents often find the right file but miss crucial lines, with line-level accuracy dropping to 14-19% in tests.
Hugging Face introduces FINAL-Bench Quantum, a benchmark suite offering five events to evaluate quantum computing methods under standardized conditions, with results from Google, IBM, and other entities included.
A new visual language model enables robots to interpret human emotional cues, improving collaboration potential.
Microsoft's SkillOpt improves GPT-5.5 by over 20 points on procedural tasks using a trained Markdown file, according to a new paper.
Google Research's Gemini-SQL2 achieved 80.04% execution accuracy on the BIRD benchmark, outperforming competitors like GPT-5.5-xhigh and Claude Opus 4.6.