List your company

Web agent task completion on WebArena Lite

0.479Success RateCATTS (H, best τ)
0.430120.442810.45550.46819Feb 12, 2026

Evaluation results

MethodLinks
CATTS (H, best τ)Gating=Entropy, N=10, Threshold (τ)=0.22026.020.479745,000
CATTS (∆, best τ)Gating=Margin, N=102026.020.479405,000
Always-arbitrateN=102026.020.44762,000
Majority voteN=102026.020.432920,000