OpenAI Refutes Exponential Time Hypothesis with New Algorithms
OpenAI's recent work challenges a long-standing assumption in computer science, sparking debate and collaboration among researchers.
387 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.
OpenAI's recent work challenges a long-standing assumption in computer science, sparking debate and collaboration among researchers.
Hugging Face introduces LitPivot, a system designed to help researchers refine their ideas by dynamically contextualizing them within the literature landscape.
A collaboration between NVIDIA and the Consortium for Biomedical Research and AI in Neurodegeneration (C-BRAIN) aims to speed up Alzheimer’s research by leveraging AI tools and data-sharing infrastructure.
Researchers from New York University and the University of Texas at Austin found AI language models overuse specific linguistic patterns that make their output difficult to read.
Hugging Face introduces two new retrieval pipelines for the ViDoRe V3 pipeline leaderboard, achieving top scores using only visual page data.
Google introduces a new AI Quest, 'Polar Peaks,' designed to teach students AI literacy while exploring neuroscience. The initiative includes teacher training modules and is part of an ongoing series.
OpenAI has released three new research findings in mathematics, computer science, and physics, with varying release strategies for each. The results include a proof of the local smoothing conjecture and the refutation of the Exponential Time Hypothesis.
Jack Clark updates his timeline for recursive self-improvement in AI, citing rapid progress in research capabilities.
A study shows AI improves immediate performance for junior lawyers but may reduce their ability to retain skills over time.
A study finds that new coding agent harnesses like Claude Code and OpenCode don't significantly boost results but can triple costs. Researchers at Shandong University tested multiple models with different harnesses and found minimal performance differences.
MIT researchers observed AI agents developing their own tasks during unsupervised exploration, suggesting new learning methods for AI systems.
Hugging Face's Lattice introduces a new 4D dataset capturing real-world human-object interactions, offering structured representations for robotics and physical AI research.