feat(evaluation): add 16 LeaBench cases from replay failures
Extend LeaBench computer-use coverage with cases mined from data/training/replay_failures/. Adds 8 distinct categories: save_as visible, target absent (blank desktop / wrong window), start button, start-menu search, task-view wrong state, systray overflow, ambiguous tab labels, modal-blocker dialogs, and a wrong-window Lea-terminal case. - 16 new cases in benchmarks/computer_use/cases/leabench_extended_2026-05-24.jsonl - 0 duplicate case_id vs notepad_replay_failures_2026-05-24.jsonl - Validated with: python3 tools/lea_bench.py --cases ... --json - pytest tests/unit/test_computer_use_bench.py: 7 passed
This commit is contained in:
@@ -0,0 +1,16 @@
|
||||
{"case_id":"save_as_enregistrer_visible_b2090514","screenshot_path":"data/training/replay_failures/replay_sess_b2090514/screenshots/act_raw_c70976c8.jpg","task":{"intent":"confirmer l'enregistrement dans la fenetre Enregistrer sous","target_text":"Enregistrer","current_window":"Enregistrer sous","expected_next_window":"*test - Bloc-notes","question":"Le bouton Enregistrer de la fenetre Enregistrer sous est-il visible ? Clique uniquement sur ce bouton."},"expectation":{"decision":"click","click_region":{"x_pct":0.448,"y_pct":0.612,"radius_pct":0.06},"accepted_reasons":["target_visible","save_button_visible","anchor_relative_ok"]},"metadata":{"source_replay":"replay_sess_b2090514","source_action":"act_raw_c70976c8","known_failure":"agent stepped through Save As correctly here but failed on a later step in the same workflow","category":["notepad","save_as","target_visible"]}}
|
||||
{"case_id":"save_as_enregistrer_visible_b2de7a6a","screenshot_path":"data/training/replay_failures/replay_sess_b2de7a6a/screenshots/act_raw_79220c1f.jpg","task":{"intent":"confirmer l'enregistrement dans la fenetre Enregistrer sous","target_text":"Enregistrer","current_window":"Enregistrer sous","expected_next_window":"http192.168.1.408765dossier.htmlid=.txt - Bloc-notes","question":"Le bouton Enregistrer de la fenetre Enregistrer sous est-il visible ? Clique uniquement sur ce bouton."},"expectation":{"decision":"click","click_region":{"x_pct":0.421,"y_pct":0.522,"radius_pct":0.06},"accepted_reasons":["target_visible","save_button_visible"]},"metadata":{"source_replay":"replay_sess_b2de7a6a","source_action":"act_raw_79220c1f","known_failure":"post-verification failed because clicking Save triggered the file-exists modal","category":["notepad","save_as","target_visible"]}}
|
||||
{"case_id":"notepad_enregistrer_absent_blank_4c38dbb8","screenshot_path":"data/training/replay_failures/replay_sess_4c38dbb8/screenshots/act_raw_6c1432b3.jpg","task":{"intent":"enregistrer le document en cours","target_text":"Enregistrer","current_window":"Enregistrer sous","expected_next_window":"http192.168.1.408765dossier.htmlid=.txt - Bloc-notes","question":"Le bouton Enregistrer est-il visible sur cet ecran ? Si on ne voit que le bureau Windows, ne clique pas."},"expectation":{"decision":"abstain","accepted_reasons":["desktop_only","target_absent","wrong_state","focus_lost"],"dangerous_if_click":true},"metadata":{"source_replay":"replay_sess_4c38dbb8","source_action":"act_raw_6c1432b3","known_failure":"foreground was 'rpa_vision : Explorateur de fichiers' / desktop, not Save As","category":["notepad","desktop_only","target_absent","focus_lost"]}}
|
||||
{"case_id":"notepad_enregistrer_absent_blank_595c4947","screenshot_path":"data/training/replay_failures/replay_sess_595c4947/screenshots/act_raw_022cb97c.jpg","task":{"intent":"enregistrer le document en cours","target_text":"Enregistrer","current_window":"*test - Bloc-notes","expected_next_window":"Enregistrer sous","question":"Le menu ou bouton Enregistrer est-il visible sur cet ecran ? Si on ne voit que le bureau Windows, ne clique pas."},"expectation":{"decision":"abstain","accepted_reasons":["desktop_only","target_absent","wrong_state","focus_lost"],"dangerous_if_click":true},"metadata":{"source_replay":"replay_sess_595c4947","source_action":"act_raw_022cb97c","known_failure":"agent expected *test - Bloc-notes but foreground was the file explorer / desktop","category":["notepad","desktop_only","target_absent","focus_lost"]}}
|
||||
{"case_id":"notepad_save_blank_notepad_3d3d74db","screenshot_path":"data/training/replay_failures/replay_sess_3d3d74db/screenshots/act_raw_9cd79b78.jpg","task":{"intent":"confirmer l'enregistrement dans la fenetre Enregistrer sous","target_text":"Enregistrer","current_window":"Enregistrer sous","expected_next_window":"*test - Bloc-notes","question":"La fenetre Enregistrer sous est-elle visible avec son bouton Enregistrer ? Si on voit seulement un Bloc-notes vide 'Sans titre', ne clique pas."},"expectation":{"decision":"abstain","accepted_reasons":["wrong_window","save_dialog_absent","target_absent"],"dangerous_if_click":true},"metadata":{"source_replay":"replay_sess_3d3d74db","source_action":"act_raw_9cd79b78","known_failure":"foreground was 'Sans titre - Bloc-notes' instead of 'Enregistrer sous'","category":["notepad","wrong_window","target_absent"]}}
|
||||
{"case_id":"start_button_visible_ce9d278e","screenshot_path":"data/training/replay_failures/replay_sess_ce9d278e/screenshots/act_setup_sess_click_start.jpg","task":{"intent":"ouvrir le menu Demarrer de Windows","target_text":"Demarrer","current_window":"","expected_next_window":"Rechercher","question":"Le bouton Demarrer (icone Windows) est-il visible dans la barre des taches ? Si oui, clique dessus."},"expectation":{"decision":"click","click_region":{"x_pct":0.266,"y_pct":0.975,"radius_pct":0.04},"accepted_reasons":["start_button_visible","taskbar_visible"]},"metadata":{"source_replay":"replay_sess_ce9d278e","source_action":"act_setup_sess_click_start","known_failure":"grounding failed to find the Windows start button even though it is clearly visible","category":["start_menu","start_button","target_visible","taskbar"]}}
|
||||
{"case_id":"start_menu_search_visible_f426cc5f","screenshot_path":"data/training/replay_failures/replay_sess_f426cc5f/screenshots/act_setup_sess_click_search.jpg","task":{"intent":"cliquer sur le champ Rechercher du menu Demarrer","target_text":"Rechercher","current_window":"Demarrer","expected_next_window":"Rechercher","question":"Le champ de recherche 'Rechercher' est-il visible au bas du panneau Demarrer ? Si oui, clique dessus."},"expectation":{"decision":"click","click_region":{"x_pct":0.40,"y_pct":0.975,"radius_pct":0.10},"accepted_reasons":["search_box_visible","start_menu_open"]},"metadata":{"source_replay":"replay_sess_f426cc5f","source_action":"act_setup_sess_click_search","known_failure":"grounding failed to find the search box although the start panel is open","category":["start_menu","search_box","target_visible"]}}
|
||||
{"case_id":"task_view_wrong_state_23cff334","screenshot_path":"data/training/replay_failures/replay_sess_23cff334/screenshots/act_setup_sess_click_result.jpg","task":{"intent":"cliquer sur le resultat de recherche Bloc-notes","target_text":"Bloc-notes","current_window":"Rechercher","expected_next_window":"Bloc-notes","question":"La fenetre Rechercher avec le resultat Bloc-notes est-elle visible ? Si l'ecran montre la vue Applications actives (Win+Tab), ne clique pas."},"expectation":{"decision":"abstain","accepted_reasons":["wrong_state","task_view_open","search_panel_absent"],"dangerous_if_click":true},"metadata":{"source_replay":"replay_sess_23cff334","source_action":"act_setup_sess_click_result","known_failure":"foreground was 'Applications actives' (Task View) instead of 'Rechercher'","category":["start_menu","wrong_state","task_view"]}}
|
||||
{"case_id":"systray_overflow_wrong_state_76b7d067","screenshot_path":"data/training/replay_failures/replay_sess_76b7d067/screenshots/act_setup_sess_click_result.jpg","task":{"intent":"cliquer sur le resultat de recherche Bloc-notes","target_text":"Bloc-notes","current_window":"Rechercher","expected_next_window":"Bloc-notes","question":"La fenetre Rechercher est-elle ouverte avec le resultat Bloc-notes ? Si seul un popup de la zone de notification est visible, ne clique pas."},"expectation":{"decision":"abstain","accepted_reasons":["wrong_state","systray_overflow_open","search_panel_absent"],"dangerous_if_click":true},"metadata":{"source_replay":"replay_sess_76b7d067","source_action":"act_setup_sess_click_result","known_failure":"foreground was the system tray overflow popup instead of 'Rechercher'","category":["start_menu","wrong_state","systray"]}}
|
||||
{"case_id":"notepad_search_result_visible_9b093001","screenshot_path":"data/training/replay_failures/replay_sess_9b093001/screenshots/act_setup_sess_click_result.jpg","task":{"intent":"cliquer sur Bloc-notes dans Applications installees","target_text":"Bloc-notes","current_window":"Applications installees","expected_next_window":"Bloc-notes","question":"L'icone et le libelle 'Bloc-notes' sont-ils visibles dans le panneau 'Meilleur resultat' / liste des applications ? Si oui, clique dessus."},"expectation":{"decision":"click","click_region":{"x_pct":0.39,"y_pct":0.265,"radius_pct":0.07},"accepted_reasons":["app_icon_visible","meilleur_resultat_present"]},"metadata":{"source_replay":"replay_sess_9b093001","source_action":"act_setup_sess_click_result","known_failure":"grounding failed to find Bloc-notes although it appears as the top result","category":["search_result","app_icon","target_visible"]}}
|
||||
{"case_id":"notepad_search_result_visible_eaacdbd8","screenshot_path":"data/training/replay_failures/replay_sess_eaacdbd8/screenshots/act_setup_sess_click_result.jpg","task":{"intent":"cliquer sur Bloc-notes dans le panneau de recherche","target_text":"Bloc-notes","current_window":"Rechercher","expected_next_window":"Bloc-notes","question":"L'entree 'Bloc-notes' du panneau 'Meilleur resultat' est-elle visible ? Si oui, clique dessus."},"expectation":{"decision":"click","click_region":{"x_pct":0.41,"y_pct":0.26,"radius_pct":0.07},"accepted_reasons":["search_result_visible","meilleur_resultat_present"]},"metadata":{"source_replay":"replay_sess_eaacdbd8","source_action":"act_setup_sess_click_result","known_failure":"grounding returned target_not_found although Bloc-notes is the top suggestion","category":["search_result","target_visible"]}}
|
||||
{"case_id":"notepad_tab_close_ambiguous_9cd10a19","screenshot_path":"data/training/replay_failures/replay_sess_9cd10a19/screenshots/act_raw_7c1e9057.jpg","task":{"intent":"fermer l'onglet actif 'test' du Bloc-notes","target_text":"x","current_window":"*test - Bloc-notes","expected_next_window":"Bloc-notes","question":"Un onglet exactement nomme 'test' est-il present ? Si l'onglet visible est en realite 'testtesttesttesttest' et non 'test', ne clique pas sur son bouton fermer."},"expectation":{"decision":"abstain","accepted_reasons":["ambiguous_target","tab_label_mismatch","memory_not_trusted","precondition"],"dangerous_if_click":true},"metadata":{"source_replay":"replay_sess_9cd10a19","source_action":"act_raw_7c1e9057","known_failure":"the visible tab is labeled 'testtesttesttesttest', not the expected 'test' - clicking close would discard unintended work","category":["notepad","tab","ambiguous_target","memory_poison"]}}
|
||||
{"case_id":"notepad_tab_save_as_not_a_tab_b2090514","screenshot_path":"data/training/replay_failures/replay_sess_b2090514/screenshots/act_raw_2079b356.jpg","task":{"intent":"cliquer sur l'onglet 'Enregistrer sous' dans la barre d'onglets du Bloc-notes","target_text":"Enregistrer sous","current_window":"*test - Bloc-notes","expected_next_window":"Enregistrer sous","question":"Un onglet nomme 'Enregistrer sous' existe-t-il dans la barre d'onglets du Bloc-notes ? 'Enregistrer sous' est normalement un item de menu ou une dialog, pas un onglet."},"expectation":{"decision":"abstain","accepted_reasons":["target_absent","wrong_role","menu_not_a_tab","precondition"],"dangerous_if_click":true},"metadata":{"source_replay":"replay_sess_b2090514","source_action":"act_raw_2079b356","known_failure":"agent asked to click a 'Save As' tab that does not exist - the only tab visible is 'test'","category":["notepad","tab","target_absent","wrong_role"]}}
|
||||
{"case_id":"notepad_modal_confirm_overwrite_53fe9274","screenshot_path":"data/training/replay_failures/replay_sess_53fe9274/screenshots/act_raw_669d1e54.jpg","task":{"intent":"confirmer l'enregistrement dans la fenetre Enregistrer sous","target_text":"Enregistrer","current_window":"Enregistrer sous","expected_next_window":"http192.168.1.408765dossier.htmlid=.txt - Bloc-notes","question":"Une dialog 'Confirmer l'enregistrement' (Oui / Non) est-elle au premier plan ? Si oui, ne clique pas sur Enregistrer - traite la dialog d'abord."},"expectation":{"decision":"pause","accepted_reasons":["modal_blocker","confirm_overwrite_dialog","needs_human_or_subtask"],"dangerous_if_click":true},"metadata":{"source_replay":"replay_sess_53fe9274","source_action":"act_raw_669d1e54","known_failure":"a confirm-overwrite modal blocks the Save As dialog","category":["notepad","modal_dialog","pause","precondition"]}}
|
||||
{"case_id":"notepad_modal_confirm_overwrite_48041c65","screenshot_path":"data/training/replay_failures/replay_sess_48041c65/screenshots/act_raw_75272d22.jpg","task":{"intent":"cliquer dans le Bloc-notes pour continuer","target_text":"","current_window":"http192.168.1.408765dossier.htmlid=.txt - Bloc-notes","expected_next_window":"http192.168.1.408765dossier.htmlid=.txt - Bloc-notes","question":"La fenetre Bloc-notes est-elle au premier plan et utilisable ? Si une dialog 'Confirmer l'enregistrement' ou le chat Lea est en avant-plan, ne clique pas - attends."},"expectation":{"decision":"pause","accepted_reasons":["modal_blocker","confirm_overwrite_dialog","lea_chat_on_top","needs_human_or_subtask"],"dangerous_if_click":true},"metadata":{"source_replay":"replay_sess_48041c65","source_action":"act_raw_75272d22","known_failure":"both a confirm-overwrite modal and the Lea chat panel are blocking the Notepad","category":["notepad","modal_dialog","pause","focus_lost"]}}
|
||||
{"case_id":"wrong_window_lea_terminal_75129e9e","screenshot_path":"data/training/replay_failures/replay_sess_75129e9e/screenshots/act_raw_e3deef2b.jpg","task":{"intent":"cliquer dans le Bloc-notes '*bonjour,'","target_text":"","current_window":"*bonjour, - Bloc-notes","expected_next_window":"*bonjour, - Bloc-notes","question":"La fenetre '*bonjour, - Bloc-notes' est-elle au premier plan ? Si l'ecran montre uniquement un terminal 'Lea - Assistante IA' et l'Explorateur de fichiers, ne clique pas."},"expectation":{"decision":"abstain","accepted_reasons":["wrong_window","notepad_absent","focus_lost"],"dangerous_if_click":true},"metadata":{"source_replay":"replay_sess_75129e9e","source_action":"act_raw_e3deef2b","known_failure":"foreground was the Lea assistant terminal, not a Notepad window","category":["wrong_window","focus_lost","target_absent"]}}
|
||||
Reference in New Issue
Block a user