Computer Use Sandboxes for Apple Devices. Use.Computer.

An ultra long horizon benchmark. Abundant. Website | GitHub

Ensuring task quality. GitHub | [coming soon]

Coding agents even for Knowledge Work. Trainloop and Mercor. GitHub | Blog

A taxonomy of why models cheat. Abundant. GitHub | Blog

A Standard for Unifying Agent Benchmarks. ServiceNow. GitHub | Paper

Trajectory Recording and Capture of Environments. GitHub

Knowledge Graph of all Wikipedia. Stanford and Wikimedia. GitHub | [coming soon]

RL scaling on mathematical reasoning. GitHub

Knowledge graphs from arbitrary text. Stanford. GitHub | Paper