Frontier models cleared only 21 of 84 physical-world tasks in RobotWorld, a new simulation testbed posted to Hugging Face that spans manipulation, ...

A new paper from Zhuowen Tu's group proposes replacing the fixed 256-token grid used in autoregressive image models with a hierarchical quadtree.

Game, a new benchmark of 247 Godot tasks, the top coding agent tested, Opus5, reached only 50.38 out of 100 on building a game from a short ...

Decide, the Gemma-4-based 26B MoE decision model, cutting VRAM 66% from 51.1 GiB to 17.1 ...

A new paper on Hugging Face proposes Tetris3D, a 3D scene generator that is trained to make neighboring objects actually fit each other instead ...

In July, Nvidia told OpenRouter it was interested in competing for the three-year-old AI-model marketplace but needed more time to evaluate, ...

"I've never read a line of this code," reads the README for Theo Browne's new Rust port of the TypeScript 7 compiler, type checker, and language ...

Ethereum researcher Justin Drake has told the crypto industry to prepare for 'bunker mode,' arguing that AI-accelerated mathematics could break ...

Researchers from Beijing Institute of Technology and Xiaohongshu propose amortized agentic policy discovery, which trains a policy to pick ...

Researchers from Tsinghua University, Joy Future Academy / JD, and the Chinese University of Hong Kong report that in autoregressive video ...

Isomorphic Labs, the AI-powered drug-discovery company spun out of Google DeepMind, is in early talks to raise more than $2 billion at a valuation of ...

The Fudan paper reports 16x training-free length extrapolation on hybrid long-context LLMs, from a 4k pretraining window out to 64k, while keeping ...