ResearchBenchmarksWeb Browsing and Shopping on WebShop (final-step)Follow94.9Purchase RateActFocus38.42853.08967.7582.411May 14, 2026Evaluation ResultsMethodMethodLinksPurchase RateStrict Success RateActFocusAlgorithm=GRPO, Model...Algorithm=GRPO, Model scale=3B, Evaluation step=1002026.0594.951.6UniformAlgorithm=GRPO, Model...Algorithm=GRPO, Model scale=3B, Evaluation step=1002026.0540.614.1