List your company

Agreement with outcome human labels on CUAVerifierBench Browserbase OM2W (n=106)

88AccuracyUniversal Verifier (UV)
46.457.26878.8Apr 5, 2026

Evaluation results

MethodLinks
Universal Verifier (UV)Verifier=UV, Backbone=GPT-5.22026.0488650.58318
Universal VerifierVerifier=UV, Base Model=GPT-5.22026.0488650.58318
WebVoy.Verifier=WebVoy., Backbone=GPT-5.22026.0474500.361828
WebJudgeVerifier=WebJudge, Backbone=GPT-5.22026.0474460.312926
WebJudgeVerifier=WebJudge, Backbone=o4-mini2026.0464440.261240
WebJudgeVerifier=WebJudge, Base Model=o4-mini2026.0464440.261240
WebVoy.Verifier=WebVoy., Backbone=GPT-4o2026.0448350.131260
WebVoyagerVerifier=WebVoy., Base Model=GPT-4o2026.0448350.131260