The Art of Building Verifiers for Computer Use Agents
Official Schedule Context
- Date/time: 2026-07-01 · 11:40am-12:00pm
- Track/room: track TBD · Expo Stage 1 NE
- Speaker(s): Miguel González Fernández, Corby Rosset
- Session type/status: session · confirmed
Official Description
Every team building browser agents has the same problem: you can't trust your own evals. Browser
tasks are too open-ended for deterministic checks, so teams use LLM verifiers as judges, and the
judges are wrong constantly. WebVoyager misses 45% of failures. WebJudge misses 22%. Used as RL
reward, you're not training a better agent, you're training a more confident liar. This talk walks
through the Universal Verifier, open-sourced with Microsoft Research: false positive rate near zero,
Cohen's κ matching human-human agreement. Four design principles, one open benchmark, and an honest
account of where auto-research worked and where it plateaued.
Related YouTube Video
No related AI Engineer channel video found yet.
Transcript Status
No official session recording transcript was found by exact title match on the AI Engineer YouTube channel during this run.
People
Notes
- Pending transcript synthesis when an official recording or confirmed matching video is available.