1 paper · 1 filter
Jiwan Chung, JiHyuk Byun, Vibhav Vineet +1
Web agents act through long interaction sequences, yet existing benchmarks evaluate only terminal success, discarding all process information and offering little guidance on improv…