VLAarchtests3 / code /VLAarchtests2_code /VLAarchtests /artifacts /reports /ablations_v7 /ablations.md
Reveal Ablations
full_model
- mean_success: 0.280
- visibility_integral: 31.968
- corridor_availability: 0.675
- reocclusion_rate: 0.014
- disturbance_cost: 0.393
- premature_retrieve_rate: 0.125
- reocclusion_after_reveal_rate: 0.547
- planner_regret: 0.091
- chunk_commit_steps: 0.000
no_planner
- mean_success: 0.200
- visibility_integral: 6.978
- corridor_availability: 0.230
- reocclusion_rate: 0.020
- disturbance_cost: 0.110
- premature_retrieve_rate: 0.151
- reocclusion_after_reveal_rate: 0.800
- planner_regret: 0.041
- chunk_commit_steps: 0.000
- paired_paired_episodes_vs_full_model: 300.000
- paired_success_delta_vs_full_model: -0.080
- paired_visibility_delta_vs_full_model: -24.990
- paired_reocclusion_delta_vs_full_model: 0.006
- paired_disturbance_delta_vs_full_model: -0.283
no_spatial_memory
- mean_success: 0.323
- visibility_integral: 37.043
- corridor_availability: 0.825
- reocclusion_rate: 0.015
- disturbance_cost: 0.441
- premature_retrieve_rate: 0.112
- reocclusion_after_reveal_rate: 0.260
- planner_regret: 0.163
- chunk_commit_steps: 0.000
- paired_paired_episodes_vs_full_model: 300.000
- paired_success_delta_vs_full_model: 0.043
- paired_visibility_delta_vs_full_model: 5.075
- paired_reocclusion_delta_vs_full_model: 0.001
- paired_disturbance_delta_vs_full_model: 0.048
no_task_head
- mean_success: 0.280
- visibility_integral: 31.965
- corridor_availability: 0.675
- reocclusion_rate: 0.014
- disturbance_cost: 0.393
- premature_retrieve_rate: 0.125
- reocclusion_after_reveal_rate: 0.547
- planner_regret: 0.091
- chunk_commit_steps: 0.000
- paired_paired_episodes_vs_full_model: 300.000
- paired_success_delta_vs_full_model: 0.000
- paired_visibility_delta_vs_full_model: -0.003
- paired_reocclusion_delta_vs_full_model: 0.000
- paired_disturbance_delta_vs_full_model: 0.000
no_geometry
- mean_success: 0.270
- visibility_integral: 32.415
- corridor_availability: 0.675
- reocclusion_rate: 0.013
- disturbance_cost: 0.398
- premature_retrieve_rate: 0.127
- reocclusion_after_reveal_rate: 0.550
- planner_regret: 0.091
- chunk_commit_steps: 0.000
- paired_paired_episodes_vs_full_model: 300.000
- paired_success_delta_vs_full_model: -0.010
- paired_visibility_delta_vs_full_model: 0.446
- paired_reocclusion_delta_vs_full_model: -0.001
- paired_disturbance_delta_vs_full_model: 0.004
no_camera_pose
- mean_success: 0.293
- visibility_integral: 31.640
- corridor_availability: 0.681
- reocclusion_rate: 0.017
- disturbance_cost: 0.389
- premature_retrieve_rate: 0.126
- reocclusion_after_reveal_rate: 0.543
- planner_regret: 0.092
- chunk_commit_steps: 0.000
- paired_paired_episodes_vs_full_model: 300.000
- paired_success_delta_vs_full_model: 0.013
- paired_visibility_delta_vs_full_model: -0.329
- paired_reocclusion_delta_vs_full_model: 0.003
- paired_disturbance_delta_vs_full_model: -0.004