Cua
← All updates

Week of April 6, 2026

Cua-Bench

  • Tinker RL integration — cua-bench now includes a Tinker-backed path for reinforcement-learning experiments.
  • Training workflow handoff — Benchmark tasks can be used from Tinker experiments, so outcomes from cua-bench can feed an RL run.

Get every update: follow with RSS.

Open in your browser to continue

Sign-in often fails inside this app's built-in browser. Open this page in Safari or Chrome to continue.

Continue anyway