RecipeDeleteMultipleRecipesWithNoise AndroidWorld, 2026-11-09. Agent: Lusor, model: DeepSeek V4.1 Flash (low). GOAL Delete the following recipes from Broccoli app: Avocado Toast with Egg, Classic Margherita Pizza, Thai Peanut Noodle Salad. RESULT - Mobilerun's grader: pass - Google's grader: pass - Lusor time: 108.2 s; Mobilerun published time: 474.7 s (4.4x faster) - Lusor model cost: $0.0238; Mobilerun estimated cost: ~$1.30 - Lusor tool calls: 21; Mobilerun published actions: 17 - App skills: read broccoli; none has rules for this task FINAL REPORT I deleted all three recipes from Broccoli: Avocado Toast with Egg, Classic Margherita Pizza, and Thai Peanut Noodle Salad. After each deletion the app confirmed it, and the recipe list now holds 29 recipes with none of the three remaining. STEPS (21; seconds into the video) 1. 30.01-30.01 s read ok Read Broccoli skill guidance 2. 31.19-31.21 s get_ui_state ok Observe current screen, launch Broccoli if needed 3. 32.30-36.24 s launch ok Launch Broccoli app and view recipe list 4. 37.34-64.62 s map_app ok Map the recipe list to find the three target recipes 5. 66.12-66.55 s tap_element ok Search for Avocado Toast with Egg 6. 67.56-71.08 s type ok Type search query for Avocado Toast with Egg 7. 72.36-73.62 s tap_element ok Open Avocado Toast with Egg recipe 8. 74.72-75.74 s tap_element ok Open More options menu on Avocado Toast with Egg to find Delete 9. 76.54-77.46 s tap_element ok Tap Delete for Avocado Toast with Egg 10. 78.29-79.41 s tap_element ok Confirm delete, then search for Classic Margherita Pizza 11. 80.54-84.19 s type ok Search for Classic Margherita Pizza 12. 85.12-86.81 s tap_element ok Open Classic Margherita Pizza and delete it 13. 88.47-89.22 s tap_element ok Tap More options on Classic Margherita Pizza 14. 90.26-91.21 s tap_element ok Tap Delete menu item for Classic Margherita Pizza 15. 92.32-93.52 s tap_element ok Confirm delete, then search for Thai Peanut Noodle Salad 16. 94.67-102.93 s chain ok Search and delete Thai Peanut Noodle Salad 17. 104.21-104.92 s tap_element ok Close search and recount list with map_app levels 1 18. 106.04-106.06 s map_app ok Recount full list with map_app levels 1 to confirm 29 recipes and no targets 19. 107.14-107.45 s key ok Close keyboard, then map_app levels 1 to recount 20. 108.95-132.37 s map_app ok Recount full list to confirm 29 recipes with no targets 21. 134.61-134.62 s done ok Task complete FILES - Video: video.mp4 (H.264), video.webm (VP9); poster: poster.webp - Data: task.json, steps.json