6 articles
Researchers from Tencent and Tsinghua University have introduced CALM, a revolutionary AI framework that ditches traditional token-by-token prediction for continuous vector reasoning. Early results show 44% less training compute and 4× faster generation, potentially reshaping how language models work.
Meta introduced SPICE, a self-play system that lets AI models learn from real-world data without human supervision—delivering major improvements in reasoning accuracy and beating all previous self-improvement methods.
Researchers have unveiled DeepAgent—a breakthrough reasoning model that can think, discover tools, and act on its own—outperforming GPT-4-level agents across major AI benchmarks.
Baidu's new PaddleOCR-VL model proves that smaller can be smarter, delivering state-of-the-art document understanding with just 0.9 billion parameters while outperforming industry giants.
We use cookies to improve your experience on our site and to show you relevant advertising. To find our more, read our privacy policy and cookie policy
Marina Lyubimova
Peter Smith
Alex Dudov
Saad Ullah
