From 7b7d7a08a5e78c3379397f01f0c9af97d5d84fbd Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Mon, 10 Aug 2026 10:34:53 -0500 Subject: results: select four-step PickupLoc history --- results/babyai_shared/pickup_p1_selector.json | 53 +++++++++++++++++++++++++++ 1 file changed, 53 insertions(+) create mode 100644 results/babyai_shared/pickup_p1_selector.json (limited to 'results/babyai_shared/pickup_p1_selector.json') diff --git a/results/babyai_shared/pickup_p1_selector.json b/results/babyai_shared/pickup_p1_selector.json new file mode 100644 index 0000000..2f5fc07 --- /dev/null +++ b/results/babyai_shared/pickup_p1_selector.json @@ -0,0 +1,53 @@ +{ + "candidates": [ + { + "bp_action_accuracy_percent": 87.26006317324047, + "bp_mission_lesion_drop_points": 33.00000000000001, + "bp_mission_lesion_success_percent": 44.6, + "bp_rollout_success_percent": 77.60000000000001, + "checks": { + "both_finite": true, + "bp_mission_lesion_drop_at_least_20": true, + "bp_success_at_least_70": true, + "clean_kp_success_at_least_60": true + }, + "clean_kp_action_accuracy_percent": 85.78602089576415, + "clean_kp_rollout_success_percent": 61.6, + "eligible": true, + "history_steps": 4 + }, + { + "bp_action_accuracy_percent": 86.68502470235686, + "bp_mission_lesion_drop_points": 29.800000000000004, + "bp_mission_lesion_success_percent": 37.4, + "bp_rollout_success_percent": 67.2, + "checks": { + "both_finite": true, + "bp_mission_lesion_drop_at_least_20": true, + "bp_success_at_least_70": false, + "clean_kp_success_at_least_60": false + }, + "clean_kp_action_accuracy_percent": 83.34818174455333, + "clean_kp_rollout_success_percent": 42.199999999999996, + "eligible": false, + "history_steps": 8 + } + ], + "gate": "pass", + "raw_or_sdil_results_run_or_read": false, + "selected": { + "context_gain": 1.0, + "hidden_layers": 4, + "history_steps": 4, + "learning_rate": 0.03, + "p2_epochs": 40, + "p2_model_and_shuffle_seeds": [ + 4101, + 4102, + 4103 + ], + "width": 256 + }, + "stage": "babyai_pickup_p1_history_selector", + "test_split_generated_or_read": false +} -- cgit v1.2.3