cua-speedrun goes open: leaderboard, paper, and code from CMU
kohjingyu · x · 2026-10-01
The cua-speedrun project was led by Jing Yu Koh with Pranjal Aggarwal and Lawrence Jang at CMU's SCS, with Sean Welleck, Daniel Fried, and Ruslan Salakhutdinov. The team invites the community to push the speed frontier of computer-use agents and has released the interactive leaderboard, paper, and code.
Related event: cua-speedrun Goes Open Source: Leaderboard, Paper, and Code Released(2 posts)→
More from Research
- FW-Icon: An Expert-Graded Benchmark for SVG Icon Creativity and Execution — andpoul · 2026-10-02
- NYU's DAGent: evaluate-then-grow planning lifts deep research agents by up to 5.8 points — newyorkuniversity · 2026-10-02
- Amazon's MILO auto-evolves agent harnesses, hitting 86.1% on Terminal-Bench 2.1 — amazon · 2026-10-02
- Prompted identity splits multi-agent LLM teams: success drops from 96% to 81% — Xavier Del Giudice · 2026-10-02
- Microsoft's Rubric Response Theory beats GRPO by up to 5.6 points with half the judge budget — microsoft · 2026-10-02
- New SI-cpWER benchmark tests persistent speaker identity across meetings; ThyVoice leads at 47.13 — Shantanu Vispute · 2026-10-02