Add files using upload-large-folder tool

artifacts/outputs/interaction/proxy_interaction_state/config_resolved.yaml

@@ -0,0 +1,120 @@
+experiment_name: proxy_interaction_state
+output_dir: /workspace/VLAarchtests/artifacts/outputs/interaction
+device: cuda
+seed: 13
+data:
+  proxies:
+  - foliage_proxy
+  - bag_proxy
+  - cloth_proxy
+  resolution: 96
+  train_episodes_per_proxy: 48
+  val_episodes_per_proxy: 16
+  train_dataset_path: /workspace/VLAarchtests/artifacts/data/reveal_proxy/proxy_train_v4_noleak_counterfactual.pt
+  val_dataset_path: /workspace/VLAarchtests/artifacts/data/reveal_proxy/proxy_val_v4_noleak_counterfactual.pt
+  rebuild_dataset: false
+  chunk_horizon: 8
+  rollout_horizon: 5
+  history_steps: 6
+  planner_candidates: 8
+  seed: 13
+optim:
+  epochs: 8
+  batch_size: 16
+  num_workers: 0
+  lr: 0.001
+  weight_decay: 0.0001
+trainer:
+  policy_type: interaction_state
+  use_bf16: true
+  grad_clip_norm: 1.0
+  freeze_backbone: true
+  gradient_checkpointing: false
+  plan_during_train: true
+  plan_during_eval: true
+  support_mode_conditioning: true
+  planner_mode: trainable
+policy:
+  backbone:
+    model_name: openai/clip-vit-base-patch32
+    hidden_dim: 128
+    max_text_tokens: 32
+    freeze_backbone: true
+    gradient_checkpointing: false
+    use_dummy_backbone: true
+  fusion:
+    hidden_dim: 128
+    num_cameras: 3
+    num_layers: 2
+    num_heads: 4
+    ff_dim: 256
+    dropout: 0.1
+    proprio_dim: 32
+    proprio_tokens: 1
+  memory:
+    hidden_dim: 128
+    history_steps: 6
+    num_layers: 2
+    dropout: 0.1
+    memory_bank_size: 4
+    num_heads: 4
+    max_history_steps: 8
+  decoder:
+    hidden_dim: 128
+    num_heads: 4
+    num_layers: 2
+    ff_dim: 256
+    dropout: 0.1
+    chunk_size: 8
+    action_dim: 14
+    arm_action_dim: 7
+    num_candidates: 8
+    num_phases: 5
+    num_arm_roles: 4
+  reveal_head:
+    hidden_dim: 128
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 5
+    belief_map_size: 32
+    field_size: 16
+    num_heads: 4
+    predict_belief_map: true
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+  world_model:
+    hidden_dim: 128
+    action_dim: 14
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 5
+    field_size: 16
+    num_heads: 4
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+  planner:
+    hidden_dim: 128
+    num_candidates: 8
+    action_dim: 14
+    num_support_modes: 3
+    utility_margin: 0.1
+    num_heads: 4
+    num_layers: 2
+    num_phases: 5
+    num_arm_roles: 4
+loss_weights:
+  action: 1.0
+  support_mode: 0.15
+  corridor: 0.2
+  persistence: 0.1
+  disturbance: 0.1
+  world_model: 0.25
+  belief: 0.05
+  planner_success: 0.2
+  planner_risk: 0.1
+  planner_ranking: 0.1
+  proposal_reconstruction: 0.2
+  proposal_success: 0.1
+  proposal_ranking: 0.1

artifacts/outputs/interaction/proxy_interaction_state/diagnostics/proxy_diagnostics.json

@@ -0,0 +1,7 @@
+{
+  "planner_top1_accuracy": 0.42857142857142855,
+  "planner_regret": 0.1302730292081833,
+  "risk_calibration_mse": 0.010318642482161522,
+  "role_collapse_rate": 1.0,
+  "num_samples": 126
+}

artifacts/outputs/interaction/proxy_interaction_state/metrics.json

@@ -0,0 +1,314 @@
+[
+  {
+    "epoch": 0,
+    "train": {
+      "action": 0.08913872321136296,
+      "belief": 0.2182194758206606,
+      "corridor": 0.3059119439373414,
+      "disturbance": 0.026565797566339217,
+      "persistence": 4.247902105251948,
+      "planner_ranking": 0.1385585286964973,
+      "planner_risk": 0.01735761243617162,
+      "planner_success": 0.6056295608480772,
+      "proposal_ranking": 0.17030407022684813,
+      "proposal_reconstruction": 0.1340707956502835,
+      "proposal_success": 0.6101961359381676,
+      "reocclusion": 0.6835535218318304,
+      "support_mode": 0.7136615564425787,
+      "total": 1.6980353395144145,
+      "uncertainty": 0.25639228243380785,
+      "world_model": 2.977962007125219
+    },
+    "val": {
+      "action": 0.026339974952861667,
+      "belief": 0.12528764735907316,
+      "corridor": 0.2495101224631071,
+      "disturbance": 0.0031769238121341914,
+      "persistence": 4.169311195611954,
+      "planner_ranking": 0.10334245767444372,
+      "planner_risk": 0.011477399850264192,
+      "planner_success": 0.5406695902347565,
+      "proposal_ranking": 0.12281709257513285,
+      "proposal_reconstruction": 0.05566257517784834,
+      "proposal_success": 0.6215158998966217,
+      "reocclusion": 0.66676115244627,
+      "support_mode": 0.6648052111268044,
+      "total": 1.4975968450307846,
+      "uncertainty": 0.11675945483148098,
+      "world_model": 2.713745206594467
+    }
+  },
+  {
+    "epoch": 1,
+    "train": {
+      "action": 0.026542372380693752,
+      "belief": 0.11461626676221688,
+      "corridor": 0.24775489792227745,
+      "disturbance": 0.0027499570617995537,
+      "persistence": 4.069600621859233,
+      "planner_ranking": 0.13798692263662815,
+      "planner_risk": 0.012379237974528223,
+      "planner_success": 0.5266842084626356,
+      "proposal_ranking": 0.10805700657268365,
+      "proposal_reconstruction": 0.05344949491942922,
+      "proposal_success": 0.6047845358649889,
+      "reocclusion": 0.6740957250197729,
+      "support_mode": 0.6921681016683578,
+      "total": 1.4517146944999695,
+      "uncertainty": 0.04957114780942599,
+      "world_model": 2.5700205266475677
+    },
+    "val": {
+      "action": 0.02145913220010698,
+      "belief": 0.10869172494858503,
+      "corridor": 0.23724456690251827,
+      "disturbance": 0.0021989822344039567,
+      "persistence": 3.844908654689789,
+      "planner_ranking": 0.12535850331187248,
+      "planner_risk": 0.01136385987047106,
+      "planner_success": 0.5217581205070019,
+      "proposal_ranking": 0.10524471942335367,
+      "proposal_reconstruction": 0.04592850711196661,
+      "proposal_success": 0.6284073293209076,
+      "reocclusion": 0.6619901061058044,
+      "support_mode": 0.7111924216151237,
+      "total": 1.4133924543857574,
+      "uncertainty": 0.025329279713332653,
+      "world_model": 2.534369081258774
+    }
+  },
+  {
+    "epoch": 2,
+    "train": {
+      "action": 0.024407408433035016,
+      "belief": 0.1138611612841487,
+      "corridor": 0.2458544857800007,
+      "disturbance": 0.004206997211440466,
+      "persistence": 3.9157366255919137,
+      "planner_ranking": 0.11510052252560854,
+      "planner_risk": 0.011586199165321887,
+      "planner_success": 0.5286323353648186,
+      "proposal_ranking": 0.10212837501118581,
+      "proposal_reconstruction": 0.04920657336091002,
+      "proposal_success": 0.6089240709940592,
+      "reocclusion": 0.6643339196840922,
+      "support_mode": 0.6847635904947916,
+      "total": 1.3605846464633942,
+      "uncertainty": 0.02807092812145129,
+      "world_model": 2.2947811782360077
+    },
+    "val": {
+      "action": 0.02334526553750038,
+      "belief": 0.15983260795474052,
+      "corridor": 0.28226315788924694,
+      "disturbance": 0.017097398158512078,
+      "persistence": 3.902952551841736,
+      "planner_ranking": 0.10127842240035534,
+      "planner_risk": 0.012675322766881436,
+      "planner_success": 0.5396927110850811,
+      "proposal_ranking": 0.10627106577157974,
+      "proposal_reconstruction": 0.04857080662623048,
+      "proposal_success": 0.6166459769010544,
+      "reocclusion": 0.6506054624915123,
+      "support_mode": 0.665102630853653,
+      "total": 1.2038391679525375,
+      "uncertainty": 0.08320197206921875,
+      "world_model": 1.6363814026117325
+    }
+  },
+  {
+    "epoch": 3,
+    "train": {
+      "action": 0.02514208061620593,
+      "belief": 0.11728278330216806,
+      "corridor": 0.24663691284755865,
+      "disturbance": 0.006238660357970123,
+      "persistence": 3.951133112112681,
+      "planner_ranking": 0.1094918263455232,
+      "planner_risk": 0.010860085254535079,
+      "planner_success": 0.4831004776060581,
+      "proposal_ranking": 0.09784856469680865,
+      "proposal_reconstruction": 0.04991315118968487,
+      "proposal_success": 0.6049580449859301,
+      "reocclusion": 0.6561904201904932,
+      "support_mode": 0.6671389614542326,
+      "total": 1.2903279686967533,
+      "uncertainty": 0.028790034314927954,
+      "world_model": 2.047424097855886
+    },
+    "val": {
+      "action": 0.026369616854935884,
+      "belief": 0.10613519884645939,
+      "corridor": 0.22941571101546288,
+      "disturbance": 0.003113830098300241,
+      "persistence": 3.742612451314926,
+      "planner_ranking": 0.09627582784742117,
+      "planner_risk": 0.01076946739340201,
+      "planner_success": 0.47833504155278206,
+      "proposal_ranking": 0.09593241196125746,
+      "proposal_reconstruction": 0.0534784197807312,
+      "proposal_success": 0.6306303292512894,
+      "reocclusion": 0.6404713988304138,
+      "support_mode": 0.6704101711511612,
+      "total": 1.1889225989580154,
+      "uncertainty": 0.02097803057404235,
+      "world_model": 1.7339447140693665
+    }
+  },
+  {
+    "epoch": 4,
+    "train": {
+      "action": 0.02463891791800658,
+      "belief": 0.10698648542165756,
+      "corridor": 0.23474377828339735,
+      "disturbance": 0.0030594476411351934,
+      "persistence": 3.8452726304531097,
+      "planner_ranking": 0.10481727588921785,
+      "planner_risk": 0.011097095673903823,
+      "planner_success": 0.44664503013094264,
+      "proposal_ranking": 0.09467410761862993,
+      "proposal_reconstruction": 0.04960497561842203,
+      "proposal_success": 0.6051262418429056,
+      "reocclusion": 0.6540075515707334,
+      "support_mode": 0.6753946940104166,
+      "total": 1.2287767976522446,
+      "uncertainty": 0.02127678649655233,
+      "world_model": 1.8863319009542465
+    },
+    "val": {
+      "action": 0.019384633051231503,
+      "belief": 0.10288739763200283,
+      "corridor": 0.22519288212060928,
+      "disturbance": 0.0021419554541353136,
+      "persistence": 3.7146196961402893,
+      "planner_ranking": 0.10946678463369608,
+      "planner_risk": 0.011642372934147716,
+      "planner_success": 0.47963740676641464,
+      "proposal_ranking": 0.08971122652292252,
+      "proposal_reconstruction": 0.04226163262501359,
+      "proposal_success": 0.6239738017320633,
+      "reocclusion": 0.6498126462101936,
+      "support_mode": 0.6662197485566139,
+      "total": 1.228643923997879,
+      "uncertainty": 0.010205027123447508,
+      "world_model": 1.946038767695427
+    }
+  },
+  {
+    "epoch": 5,
+    "train": {
+      "action": 0.023295348898197215,
+      "belief": 0.10408696439117193,
+      "corridor": 0.23216054091850916,
+      "disturbance": 0.002513651920404906,
+      "persistence": 3.805388251940409,
+      "planner_ranking": 0.09554585038373868,
+      "planner_risk": 0.010421493653363237,
+      "planner_success": 0.438693817704916,
+      "proposal_ranking": 0.09420504545172055,
+      "proposal_reconstruction": 0.04733991126219431,
+      "proposal_success": 0.6043872957428297,
+      "reocclusion": 0.6555871988336245,
+      "support_mode": 0.6715305869777998,
+      "total": 1.1565892845392227,
+      "uncertainty": 0.0161813710971425,
+      "world_model": 1.636805718143781
+    },
+    "val": {
+      "action": 0.020826205611228943,
+      "belief": 0.10143636167049408,
+      "corridor": 0.2232594806700945,
+      "disturbance": 0.001781191662303172,
+      "persistence": 3.6273659467697144,
+      "planner_ranking": 0.16842342726886272,
+      "planner_risk": 0.01012508076382801,
+      "planner_success": 0.5004390366375446,
+      "proposal_ranking": 0.08787750266492367,
+      "proposal_reconstruction": 0.04392921878024936,
+      "proposal_success": 0.624223992228508,
+      "reocclusion": 0.6372809633612633,
+      "support_mode": 0.664175495505333,
+      "total": 1.1100417152047157,
+      "uncertainty": 0.01541757385712117,
+      "world_model": 1.4644491225481033
+    }
+  },
+  {
+    "epoch": 6,
+    "train": {
+      "action": 0.02248762206484874,
+      "belief": 0.10121824126690626,
+      "corridor": 0.2321634783099095,
+      "disturbance": 0.0020452279277378693,
+      "persistence": 3.836012969414393,
+      "planner_ranking": 0.11302779614925385,
+      "planner_risk": 0.01022351956150184,
+      "planner_success": 0.4348009414970875,
+      "proposal_ranking": 0.09360451406488816,
+      "proposal_reconstruction": 0.0462404570231835,
+      "proposal_success": 0.6066909035046896,
+      "reocclusion": 0.6465126698215803,
+      "support_mode": 0.6658426920572916,
+      "total": 1.1116433491309483,
+      "uncertainty": 0.013347609317861497,
+      "world_model": 1.4494130065043767
+    },
+    "val": {
+      "action": 0.01996314455755055,
+      "belief": 0.10138983093202114,
+      "corridor": 0.22823662497103214,
+      "disturbance": 0.002091344154905528,
+      "persistence": 3.6982297003269196,
+      "planner_ranking": 0.09198376163840294,
+      "planner_risk": 0.010279824957251549,
+      "planner_success": 0.47799063101410866,
+      "proposal_ranking": 0.09706298727542162,
+      "proposal_reconstruction": 0.04272295255213976,
+      "proposal_success": 0.6218197643756866,
+      "reocclusion": 0.6380461752414703,
+      "support_mode": 0.6696624308824539,
+      "total": 1.0816360414028168,
+      "uncertainty": 0.007035618589725345,
+      "world_model": 1.3655437678098679
+    }
+  },
+  {
+    "epoch": 7,
+    "train": {
+      "action": 0.021682963396112125,
+      "belief": 0.10601237788796425,
+      "corridor": 0.23365325729052225,
+      "disturbance": 0.002228843019111082,
+      "persistence": 3.7541938523451486,
+      "planner_ranking": 0.0957260721673568,
+      "planner_risk": 0.01018386852229014,
+      "planner_success": 0.4461444852252801,
+      "proposal_ranking": 0.09633061879624923,
+      "proposal_reconstruction": 0.044885387644171715,
+      "proposal_success": 0.6029221788048744,
+      "reocclusion": 0.6449378555019697,
+      "support_mode": 0.6726728156208992,
+      "total": 1.0910143380363782,
+      "uncertainty": 0.01619089951661105,
+      "world_model": 1.3958957940340042
+    },
+    "val": {
+      "action": 0.01988930464722216,
+      "belief": 0.10534531436860561,
+      "corridor": 0.22113148868083954,
+      "disturbance": 0.001838716707425192,
+      "persistence": 3.629800319671631,
+      "planner_ranking": 0.12714975513517857,
+      "planner_risk": 0.008069174771662802,
+      "planner_success": 0.6245891973376274,
+      "proposal_ranking": 0.10471091698855162,
+      "proposal_reconstruction": 0.043083535972982645,
+      "proposal_success": 0.6201159805059433,
+      "reocclusion": 0.6348166316747665,
+      "support_mode": 0.6840479746460915,
+      "total": 1.1184726804494858,
+      "uncertainty": 0.022346424055285752,
+      "world_model": 1.4034390151500702
+    }
+  }
+]

artifacts/outputs/interaction/proxy_interaction_state_smoke/config_resolved.yaml

@@ -0,0 +1,120 @@
+experiment_name: proxy_interaction_state_smoke
+output_dir: /workspace/VLAarchtests/artifacts/outputs/interaction
+device: cuda
+seed: 13
+data:
+  proxies:
+  - foliage_proxy
+  - bag_proxy
+  - cloth_proxy
+  resolution: 64
+  train_episodes_per_proxy: 6
+  val_episodes_per_proxy: 2
+  train_dataset_path: /workspace/VLAarchtests/artifacts/data/reveal_proxy/proxy_train_smoke_v4.pt
+  val_dataset_path: /workspace/VLAarchtests/artifacts/data/reveal_proxy/proxy_val_smoke_v4.pt
+  rebuild_dataset: false
+  chunk_horizon: 4
+  rollout_horizon: 3
+  history_steps: 6
+  planner_candidates: 4
+  seed: 13
+optim:
+  epochs: 2
+  batch_size: 8
+  num_workers: 0
+  lr: 0.001
+  weight_decay: 0.0001
+trainer:
+  policy_type: interaction_state
+  use_bf16: true
+  grad_clip_norm: 1.0
+  freeze_backbone: true
+  gradient_checkpointing: false
+  plan_during_train: true
+  plan_during_eval: true
+  support_mode_conditioning: true
+  planner_mode: trainable
+policy:
+  backbone:
+    model_name: openai/clip-vit-base-patch32
+    hidden_dim: 64
+    max_text_tokens: 32
+    freeze_backbone: true
+    gradient_checkpointing: false
+    use_dummy_backbone: true
+  fusion:
+    hidden_dim: 64
+    num_cameras: 3
+    num_layers: 2
+    num_heads: 4
+    ff_dim: 128
+    dropout: 0.1
+    proprio_dim: 32
+    proprio_tokens: 1
+  memory:
+    hidden_dim: 64
+    history_steps: 6
+    num_layers: 2
+    dropout: 0.1
+    memory_bank_size: 4
+    num_heads: 4
+    max_history_steps: 8
+  decoder:
+    hidden_dim: 64
+    num_heads: 4
+    num_layers: 2
+    ff_dim: 128
+    dropout: 0.1
+    chunk_size: 4
+    action_dim: 14
+    arm_action_dim: 7
+    num_candidates: 4
+    num_phases: 5
+    num_arm_roles: 4
+  reveal_head:
+    hidden_dim: 64
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 3
+    belief_map_size: 32
+    field_size: 16
+    num_heads: 4
+    predict_belief_map: true
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+  world_model:
+    hidden_dim: 64
+    action_dim: 14
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 3
+    field_size: 16
+    num_heads: 4
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+  planner:
+    hidden_dim: 64
+    num_candidates: 4
+    action_dim: 14
+    num_support_modes: 3
+    utility_margin: 0.1
+    num_heads: 4
+    num_layers: 2
+    num_phases: 5
+    num_arm_roles: 4
+loss_weights:
+  action: 1.0
+  support_mode: 0.15
+  corridor: 0.2
+  persistence: 0.1
+  disturbance: 0.1
+  world_model: 0.25
+  belief: 0.05
+  planner_success: 0.2
+  planner_risk: 0.1
+  planner_ranking: 0.1
+  proposal_reconstruction: 0.2
+  proposal_success: 0.1
+  proposal_ranking: 0.1

artifacts/outputs/interaction/proxy_interaction_state_smoke/diagnostics/proxy_diagnostics.json

@@ -0,0 +1,7 @@
+{
+  "planner_top1_accuracy": 0.375,
+  "planner_regret": 0.4054882526397705,
+  "risk_calibration_mse": 0.01106204278767109,
+  "role_collapse_rate": 1.0,
+  "num_samples": 16
+}

artifacts/outputs/interaction/proxy_interaction_state_smoke/metrics.json

@@ -0,0 +1,88 @@
+[
+  {
+    "epoch": 0,
+    "train": {
+      "action": 0.15396334496991976,
+      "arm_role": 1.0464913504464286,
+      "belief": 0.5699181130954197,
+      "corridor": 0.5832605191639492,
+      "disturbance": 0.10345468350819179,
+      "persistence": 1.7334490844181605,
+      "phase": 1.1630859375,
+      "planner_ranking": 0.1581665085894721,
+      "planner_risk": 0.02096142333799175,
+      "planner_success": 0.6495738455227443,
+      "proposal_ranking": 0.19803496769496373,
+      "proposal_reconstruction": 0.20699733815022878,
+      "proposal_success": 0.6321364641189575,
+      "reocclusion": 0.6904009750911168,
+      "support_mode": 0.8394252232142857,
+      "total": 1.6638009548187256,
+      "uncertainty": 0.6862367732184274,
+      "world_model": 2.6067678928375244
+    },
+    "val": {
+      "action": 0.06714679300785065,
+      "arm_role": 0.8156493306159973,
+      "belief": 0.4122387021780014,
+      "corridor": 0.43514861166477203,
+      "disturbance": 0.04421420395374298,
+      "persistence": 1.679144263267517,
+      "phase": 0.795434981584549,
+      "planner_ranking": 0.10125493630766869,
+      "planner_risk": 0.014165664557367563,
+      "planner_success": 0.7003250122070312,
+      "proposal_ranking": 0.19173914939165115,
+      "proposal_reconstruction": 0.10236449539661407,
+      "proposal_success": 0.6280835568904877,
+      "reocclusion": 0.6901499629020691,
+      "support_mode": 0.7109985053539276,
+      "total": 1.3141186237335205,
+      "uncertainty": 0.560584306716919,
+      "world_model": 2.025277078151703
+    }
+  },
+  {
+    "epoch": 1,
+    "train": {
+      "action": 0.06967730287994657,
+      "arm_role": 0.7705426897321429,
+      "belief": 0.3012138817991529,
+      "corridor": 0.35292174773556845,
+      "disturbance": 0.01887136205498661,
+      "persistence": 1.7558837277548653,
+      "phase": 0.8579799107142857,
+      "planner_ranking": 0.13363037151949747,
+      "planner_risk": 0.010756842027019178,
+      "planner_success": 0.6408852168491909,
+      "proposal_ranking": 0.12252488732337952,
+      "proposal_reconstruction": 0.10212650469371251,
+      "proposal_success": 0.619073646409171,
+      "reocclusion": 0.7055942841938564,
+      "support_mode": 0.7973284040178571,
+      "total": 1.2046506234577723,
+      "uncertainty": 0.47724126066480366,
+      "world_model": 1.6589686359677995
+    },
+    "val": {
+      "action": 0.05438657104969025,
+      "arm_role": 0.727223664522171,
+      "belief": 0.16894277185201645,
+      "corridor": 0.2628409117460251,
+      "disturbance": 0.0024247884866781533,
+      "persistence": 1.7001646757125854,
+      "phase": 0.7108556628227234,
+      "planner_ranking": 0.10945596918463707,
+      "planner_risk": 0.010194436646997929,
+      "planner_success": 0.6684828102588654,
+      "proposal_ranking": 0.09799150377511978,
+      "proposal_reconstruction": 0.09263791516423225,
+      "proposal_success": 0.6696180105209351,
+      "reocclusion": 0.7087761461734772,
+      "support_mode": 0.6864266693592072,
+      "total": 1.1171032786369324,
+      "uncertainty": 0.3682785928249359,
+      "world_model": 1.591063678264618
+    }
+  }
+]

artifacts/outputs/interaction/proxy_reveal_state_full_diagnostics/proxy_diagnostics.json

@@ -0,0 +1,7 @@
+{
+  "planner_top1_accuracy": 0.2698412698412698,
+  "planner_regret": 0.3642531931400299,
+  "risk_calibration_mse": 0.011263838969171047,
+  "role_collapse_rate": 0.0,
+  "num_samples": 126
+}

artifacts/outputs/interaction/proxy_reveal_state_smoke_diagnostics/proxy_diagnostics.json

@@ -0,0 +1,7 @@
+{
+  "planner_top1_accuracy": 0.25,
+  "planner_regret": 0.2715669870376587,
+  "risk_calibration_mse": 0.0521860271692276,
+  "role_collapse_rate": 0.0,
+  "num_samples": 16
+}

artifacts/outputs/interaction/reveal_ablation_interaction_ep1/ablations.json

@@ -0,0 +1,80 @@
+{
+  "full_model": {
+    "per_task_success": {
+      "foliage_proxy": 0.0,
+      "bag_proxy": 0.0,
+      "cloth_proxy": 0.0
+    },
+    "mean_success": 0.0,
+    "visibility_integral": 67.46570078531902,
+    "corridor_availability": 0.9835391044616699,
+    "reocclusion_rate": 0.0,
+    "persistence_horizon_mae": 2.9001280729969343,
+    "disturbance_cost": 0.7520796457926432
+  },
+  "no_interaction_head": {
+    "per_task_success": {
+      "foliage_proxy": 0.0,
+      "bag_proxy": 0.0,
+      "cloth_proxy": 0.0
+    },
+    "mean_success": 0.0,
+    "visibility_integral": 63.56013615926107,
+    "corridor_availability": 0.9876543283462524,
+    "reocclusion_rate": 0.0,
+    "persistence_horizon_mae": null,
+    "disturbance_cost": 0.8973869283994039
+  },
+  "no_world_model": {
+    "per_task_success": {
+      "foliage_proxy": 0.0,
+      "bag_proxy": 0.0,
+      "cloth_proxy": 0.0
+    },
+    "mean_success": 0.0,
+    "visibility_integral": 63.46492385864258,
+    "corridor_availability": 0.9876543283462524,
+    "reocclusion_rate": 0.0,
+    "persistence_horizon_mae": 2.8303335145115853,
+    "disturbance_cost": 0.900044322013855
+  },
+  "no_planner": {
+    "per_task_success": {
+      "foliage_proxy": 0.0,
+      "bag_proxy": 0.0,
+      "cloth_proxy": 0.0
+    },
+    "mean_success": 0.0,
+    "visibility_integral": 63.54209772745768,
+    "corridor_availability": 0.9876543283462524,
+    "reocclusion_rate": 0.0,
+    "persistence_horizon_mae": 2.8313964143395425,
+    "disturbance_cost": 0.8962932030359904
+  },
+  "no_role_tokens": {
+    "per_task_success": {
+      "foliage_proxy": 0.0,
+      "bag_proxy": 0.0,
+      "cloth_proxy": 0.0
+    },
+    "mean_success": 0.0,
+    "visibility_integral": 67.39010111490886,
+    "corridor_availability": 0.9835391044616699,
+    "reocclusion_rate": 0.0,
+    "persistence_horizon_mae": 2.8990648205081624,
+    "disturbance_cost": 0.7544853488604227
+  },
+  "short_history": {
+    "per_task_success": {
+      "foliage_proxy": 0.0,
+      "bag_proxy": 0.0,
+      "cloth_proxy": 0.0
+    },
+    "mean_success": 0.0,
+    "visibility_integral": 67.46627298990886,
+    "corridor_availability": 0.9835391044616699,
+    "reocclusion_rate": 0.0,
+    "persistence_horizon_mae": 2.900133213897546,
+    "disturbance_cost": 0.7520630160967509
+  }
+}

artifacts/outputs/interaction/reveal_ablation_interaction_ep1/ablations.md

@@ -0,0 +1,49 @@
+# Reveal Ablations
+
+## full_model
+- mean_success: 0.000
+- visibility_integral: 67.466
+- corridor_availability: 0.984
+- reocclusion_rate: 0.000
+- persistence_horizon_mae: 2.900
+- disturbance_cost: 0.752
+
+## no_interaction_head
+- mean_success: 0.000
+- visibility_integral: 63.560
+- corridor_availability: 0.988
+- reocclusion_rate: 0.000
+- persistence_horizon_mae: 0.000
+- disturbance_cost: 0.897
+
+## no_world_model
+- mean_success: 0.000
+- visibility_integral: 63.465
+- corridor_availability: 0.988
+- reocclusion_rate: 0.000
+- persistence_horizon_mae: 2.830
+- disturbance_cost: 0.900
+
+## no_planner
+- mean_success: 0.000
+- visibility_integral: 63.542
+- corridor_availability: 0.988
+- reocclusion_rate: 0.000
+- persistence_horizon_mae: 2.831
+- disturbance_cost: 0.896
+
+## no_role_tokens
+- mean_success: 0.000
+- visibility_integral: 67.390
+- corridor_availability: 0.984
+- reocclusion_rate: 0.000
+- persistence_horizon_mae: 2.899
+- disturbance_cost: 0.754
+
+## short_history
+- mean_success: 0.000
+- visibility_integral: 67.466
+- corridor_availability: 0.984
+- reocclusion_rate: 0.000
+- persistence_horizon_mae: 2.900
+- disturbance_cost: 0.752

artifacts/outputs/interaction/reveal_ablation_interaction_ep1/ablations.partial.json

@@ -0,0 +1,135 @@
+{
+  "checkpoint": "/workspace/VLAarchtests/artifacts/outputs/interaction/proxy_interaction_state_smoke/checkpoint_best.pt",
+  "episodes": 1,
+  "sections": {
+    "full_model": {
+      "mean_success": 0.0,
+      "visibility_integral": 67.46570078531902,
+      "corridor_availability": 0.9835391044616699,
+      "reocclusion_rate": 0.0,
+      "persistence_horizon_mae": 2.9001280729969343,
+      "disturbance_cost": 0.7520796457926432
+    },
+    "no_interaction_head": {
+      "mean_success": 0.0,
+      "visibility_integral": 63.56013615926107,
+      "corridor_availability": 0.9876543283462524,
+      "reocclusion_rate": 0.0,
+      "persistence_horizon_mae": 0.0,
+      "disturbance_cost": 0.8973869283994039
+    },
+    "no_world_model": {
+      "mean_success": 0.0,
+      "visibility_integral": 63.46492385864258,
+      "corridor_availability": 0.9876543283462524,
+      "reocclusion_rate": 0.0,
+      "persistence_horizon_mae": 2.8303335145115853,
+      "disturbance_cost": 0.900044322013855
+    },
+    "no_planner": {
+      "mean_success": 0.0,
+      "visibility_integral": 63.54209772745768,
+      "corridor_availability": 0.9876543283462524,
+      "reocclusion_rate": 0.0,
+      "persistence_horizon_mae": 2.8313964143395425,
+      "disturbance_cost": 0.8962932030359904
+    },
+    "no_role_tokens": {
+      "mean_success": 0.0,
+      "visibility_integral": 67.39010111490886,
+      "corridor_availability": 0.9835391044616699,
+      "reocclusion_rate": 0.0,
+      "persistence_horizon_mae": 2.8990648205081624,
+      "disturbance_cost": 0.7544853488604227
+    },
+    "short_history": {
+      "mean_success": 0.0,
+      "visibility_integral": 67.46627298990886,
+      "corridor_availability": 0.9835391044616699,
+      "reocclusion_rate": 0.0,
+      "persistence_horizon_mae": 2.900133213897546,
+      "disturbance_cost": 0.7520630160967509
+    }
+  },
+  "raw": {
+    "full_model": {
+      "per_task_success": {
+        "foliage_proxy": 0.0,
+        "bag_proxy": 0.0,
+        "cloth_proxy": 0.0
+      },
+      "mean_success": 0.0,
+      "visibility_integral": 67.46570078531902,
+      "corridor_availability": 0.9835391044616699,
+      "reocclusion_rate": 0.0,
+      "persistence_horizon_mae": 2.9001280729969343,
+      "disturbance_cost": 0.7520796457926432
+    },
+    "no_interaction_head": {
+      "per_task_success": {
+        "foliage_proxy": 0.0,
+        "bag_proxy": 0.0,
+        "cloth_proxy": 0.0
+      },
+      "mean_success": 0.0,
+      "visibility_integral": 63.56013615926107,
+      "corridor_availability": 0.9876543283462524,
+      "reocclusion_rate": 0.0,
+      "persistence_horizon_mae": null,
+      "disturbance_cost": 0.8973869283994039
+    },
+    "no_world_model": {
+      "per_task_success": {
+        "foliage_proxy": 0.0,
+        "bag_proxy": 0.0,
+        "cloth_proxy": 0.0
+      },
+      "mean_success": 0.0,
+      "visibility_integral": 63.46492385864258,
+      "corridor_availability": 0.9876543283462524,
+      "reocclusion_rate": 0.0,
+      "persistence_horizon_mae": 2.8303335145115853,
+      "disturbance_cost": 0.900044322013855
+    },
+    "no_planner": {
+      "per_task_success": {
+        "foliage_proxy": 0.0,
+        "bag_proxy": 0.0,
+        "cloth_proxy": 0.0
+      },
+      "mean_success": 0.0,
+      "visibility_integral": 63.54209772745768,
+      "corridor_availability": 0.9876543283462524,
+      "reocclusion_rate": 0.0,
+      "persistence_horizon_mae": 2.8313964143395425,
+      "disturbance_cost": 0.8962932030359904
+    },
+    "no_role_tokens": {
+      "per_task_success": {
+        "foliage_proxy": 0.0,
+        "bag_proxy": 0.0,
+        "cloth_proxy": 0.0
+      },
+      "mean_success": 0.0,
+      "visibility_integral": 67.39010111490886,
+      "corridor_availability": 0.9835391044616699,
+      "reocclusion_rate": 0.0,
+      "persistence_horizon_mae": 2.8990648205081624,
+      "disturbance_cost": 0.7544853488604227
+    },
+    "short_history": {
+      "per_task_success": {
+        "foliage_proxy": 0.0,
+        "bag_proxy": 0.0,
+        "cloth_proxy": 0.0
+      },
+      "mean_success": 0.0,
+      "visibility_integral": 67.46627298990886,
+      "corridor_availability": 0.9835391044616699,
+      "reocclusion_rate": 0.0,
+      "persistence_horizon_mae": 2.900133213897546,
+      "disturbance_cost": 0.7520630160967509
+    }
+  },
+  "elapsed_seconds": 56.18267846107483
+}

artifacts/outputs/interaction/reveal_eval_interaction_ep2/reveal_benchmark.json

@@ -0,0 +1,15 @@
+{
+  "interaction": {
+    "per_task_success": {
+      "foliage_proxy": 0.0,
+      "bag_proxy": 0.0,
+      "cloth_proxy": 0.0
+    },
+    "mean_success": 0.0,
+    "visibility_integral": 62.47385279337565,
+    "corridor_availability": 0.9876543283462524,
+    "reocclusion_rate": 0.0,
+    "persistence_horizon_mae": 2.8290874414145946,
+    "disturbance_cost": 0.9037613173325857
+  }
+}

artifacts/outputs/interaction/reveal_eval_interaction_ep2/reveal_benchmark.md

@@ -0,0 +1,13 @@
+# Reveal Proxy Benchmark
+
+## interaction
+- checkpoint: /workspace/VLAarchtests/artifacts/outputs/interaction/proxy_interaction_state_smoke/checkpoint_best.pt
+- mean_success: 0.000
+- visibility_integral: 62.474
+- corridor_availability: 0.988
+- reocclusion_rate: 0.000
+- persistence_horizon_mae: 2.829
+- disturbance_cost: 0.904
+- foliage_proxy_success: 0.000
+- bag_proxy_success: 0.000
+- cloth_proxy_success: 0.000

artifacts/outputs/interaction/reveal_eval_interaction_full_ep2/reveal_benchmark.json

@@ -0,0 +1,15 @@
+{
+  "interaction": {
+    "per_task_success": {
+      "foliage_proxy": 0.0,
+      "bag_proxy": 0.0,
+      "cloth_proxy": 0.0
+    },
+    "mean_success": 0.0,
+    "visibility_integral": 62.6574592590332,
+    "corridor_availability": 0.9876543283462524,
+    "reocclusion_rate": 0.0,
+    "persistence_horizon_mae": 2.1672919432322186,
+    "disturbance_cost": 0.8980477948983511
+  }
+}

artifacts/outputs/interaction/reveal_eval_interaction_full_ep2/reveal_benchmark.md

@@ -0,0 +1,13 @@
+# Reveal Proxy Benchmark
+
+## interaction
+- checkpoint: /workspace/VLAarchtests/artifacts/outputs/interaction/proxy_interaction_state/checkpoint_best.pt
+- mean_success: 0.000
+- visibility_integral: 62.657
+- corridor_availability: 0.988
+- reocclusion_rate: 0.000
+- persistence_horizon_mae: 2.167
+- disturbance_cost: 0.898
+- foliage_proxy_success: 0.000
+- bag_proxy_success: 0.000
+- cloth_proxy_success: 0.000

artifacts/outputs/interaction/rlbench_env_smoke/import_and_train_command.txt

@@ -0,0 +1,19 @@
+{
+  "camera_names": [
+    "front",
+    "wrist_left",
+    "wrist_right"
+  ],
+  "resolution": [
+    224,
+    224
+  ],
+  "global_camera": "front",
+  "resolved_upstream_root": "/workspace/third_party/peract_bimanual/peract",
+  "imports": {
+    "rlbench": "ok",
+    "pyrep": "ok",
+    "yarr": "ok"
+  }
+}
+xvfb-run -a -s '-screen 0 1280x1024x24' /workspace/envs/rlbench/bin/python train.py method=PERACT_BC rlbench.task_name=multi_3cam 'rlbench.tasks=[bimanual_push_box,bimanual_lift_ball,bimanual_dual_push_buttons,bimanual_pick_plate,bimanual_put_item_in_drawer,bimanual_put_bottle_in_fridge,bimanual_handover_item,bimanual_pick_laptop,bimanual_straighten_rope,bimanual_sweep_to_dustpan,bimanual_lift_tray,bimanual_handover_item_easy,bimanual_take_tray_out_of_oven]' rlbench.demos=100 rlbench.demo_path=/workspace/data/rlbench2 replay.path=/workspace/replays/rlbench2 framework.logdir=/workspace/logs/rlbench2 framework.training_iterations=40000 framework.gpu=0 framework.env_gpu=0 framework.start_seed=0 ddp.num_devices=1 'rlbench.cameras=[front,wrist2,wrist]' 'rlbench.camera_resolution=[224,224]'

artifacts/outputs/interaction/rlbench_env_smoke/import_smoke.json

@@ -0,0 +1,18 @@
+{
+  "camera_names": [
+    "front",
+    "wrist_left",
+    "wrist_right"
+  ],
+  "resolution": [
+    224,
+    224
+  ],
+  "global_camera": "front",
+  "resolved_upstream_root": "/workspace/third_party/peract_bimanual/peract",
+  "imports": {
+    "rlbench": "ok",
+    "pyrep": "ok",
+    "yarr": "ok"
+  }
+}

artifacts/outputs/interaction/rlbench_env_smoke/launch_reset_smoke.json

@@ -0,0 +1,24 @@
+{
+  "description": "hold the drawer with left hand and open the bottom drawer with right hand",
+  "front_rgb_shape": [
+    128,
+    128,
+    3
+  ],
+  "wrist_rgb_shape": [
+    128,
+    128,
+    3
+  ],
+  "wrist2_rgb_shape": [
+    128,
+    128,
+    3
+  ],
+  "right_pose_shape": [
+    7
+  ],
+  "left_pose_shape": [
+    7
+  ]
+}

artifacts/outputs/interaction/rlbench_env_smoke/launch_smoke_cli.json

@@ -0,0 +1,40 @@
+Launching  task_design_open_drawer.ttt
+dominant hand in open_drawer init_episode:  
+base_rotation_bounds:  
+{
+  "display": ":103",
+  "headless": true,
+  "task": "OpenDrawer",
+  "description": "hold the drawer with left hand and open the bottom drawer with right hand",
+  "front_rgb_shape": [
+    128,
+    128,
+    3
+  ],
+  "wrist_rgb_shape": [
+    128,
+    128,
+    3
+  ],
+  "wrist2_rgb_shape": [
+    128,
+    128,
+    3
+  ],
+  "right_pose_shape": [
+    7
+  ],
+  "left_pose_shape": [
+    7
+  ],
+  "stepped_arm": "right",
+  "reward": 0.0,
+  "done": false,
+  "front_rgb_shape_after_step": [
+    128,
+    128,
+    3
+  ]
+}
+Environment2Robots is shutting down
+[CoppeliaSim:loadinfo]   done.

artifacts/outputs/interaction/rlbench_env_smoke/launch_step_smoke.json

@@ -0,0 +1,10 @@
+{
+  "description": "hold the drawer with left hand and open the bottom drawer with right hand",
+  "reward": 0.0,
+  "done": false,
+  "front_rgb_shape_after_step": [
+    128,
+    128,
+    3
+  ]
+}

artifacts/outputs/interaction/rlbench_open_drawer_interaction_smoke/config_resolved.yaml

@@ -0,0 +1,126 @@
+experiment_name: rlbench_open_drawer_interaction_smoke
+output_dir: /workspace/VLAarchtests/artifacts/outputs/interaction
+device: cuda
+seed: 13
+init_checkpoint: /workspace/VLAarchtests/artifacts/outputs/interaction/proxy_interaction_state_smoke/checkpoint_best.pt
+init_strict: false
+data:
+  dataset_root: /workspace/VLAarchtests/artifacts/data/rlbench_smoke_open_drawer
+  tasks:
+  - open_drawer
+  train_episodes:
+  - 0
+  val_episodes:
+  - 1
+  resolution: 128
+  chunk_horizon: 4
+  proprio_dim: 32
+  history_steps: 6
+  max_train_samples: 8
+  max_val_samples: 4
+optim:
+  epochs: 1
+  batch_size: 2
+  num_workers: 0
+  lr: 0.0005
+  weight_decay: 0.0001
+trainer:
+  policy_type: interaction_state
+  use_bf16: true
+  grad_clip_norm: 1.0
+  freeze_backbone: true
+  gradient_checkpointing: false
+  plan_during_train: false
+  plan_during_eval: true
+  support_mode_conditioning: true
+  planner_mode: proxy_pretrained
+policy:
+  backbone:
+    model_name: openai/clip-vit-base-patch32
+    hidden_dim: 64
+    max_text_tokens: 32
+    freeze_backbone: true
+    gradient_checkpointing: false
+    use_dummy_backbone: true
+  fusion:
+    hidden_dim: 64
+    num_cameras: 3
+    num_layers: 2
+    num_heads: 4
+    ff_dim: 128
+    dropout: 0.1
+    proprio_dim: 32
+    proprio_tokens: 1
+  memory:
+    hidden_dim: 64
+    history_steps: 6
+    num_layers: 2
+    dropout: 0.1
+    memory_bank_size: 4
+    num_heads: 4
+    max_history_steps: 8
+  decoder:
+    hidden_dim: 64
+    num_heads: 4
+    num_layers: 2
+    ff_dim: 128
+    dropout: 0.1
+    chunk_size: 4
+    action_dim: 14
+    arm_action_dim: 7
+    num_candidates: 4
+    num_phases: 5
+    num_arm_roles: 4
+  reveal_head:
+    hidden_dim: 64
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 3
+    belief_map_size: 32
+    field_size: 16
+    num_heads: 4
+    predict_belief_map: true
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+  world_model:
+    hidden_dim: 64
+    action_dim: 14
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 3
+    field_size: 16
+    num_heads: 4
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+    belief_map_size: 32
+    predict_belief_map: true
+  planner:
+    hidden_dim: 64
+    num_candidates: 4
+    action_dim: 14
+    num_support_modes: 3
+    utility_margin: 0.1
+    corridor_weight: 1.0
+    persistence_weight: 0.5
+    proposal_weight: 0.5
+    task_progress_weight: 0.75
+    disturbance_weight: 0.75
+    reocclusion_weight: 0.5
+    visibility_weight: 0.25
+    num_heads: 4
+    num_layers: 2
+    num_phases: 5
+    num_arm_roles: 4
+loss_weights:
+  action: 1.0
+  support_mode: 0.0
+  corridor: 0.0
+  persistence: 0.0
+  disturbance: 0.0
+  world_model: 0.0
+  belief: 0.0
+  planner_success: 0.0
+  planner_risk: 0.0
+  planner_ranking: 0.0

artifacts/outputs/interaction/rlbench_open_drawer_interaction_smoke/metrics.json

@@ -0,0 +1,27 @@
+[
+  {
+    "epoch": 0,
+    "train": {
+      "action": 0.1628243774175644,
+      "planner_ranking": 0.0,
+      "planner_risk": 0.0,
+      "planner_success": 0.0,
+      "proposal_ranking": 0.0,
+      "proposal_reconstruction": 0.0,
+      "proposal_success": 0.0,
+      "total": 0.1628243774175644,
+      "world_model": 0.0
+    },
+    "val": {
+      "action": 0.05217807739973068,
+      "planner_ranking": 0.0,
+      "planner_risk": 0.0,
+      "planner_success": 0.0,
+      "proposal_ranking": 0.0,
+      "proposal_reconstruction": 0.0,
+      "proposal_success": 0.0,
+      "total": 0.05217807739973068,
+      "world_model": 0.0
+    }
+  }
+]

artifacts/outputs/interaction/rlbench_open_drawer_interaction_smoke/summary.json

@@ -0,0 +1,50 @@
+{
+  "experiment_name": "rlbench_open_drawer_interaction_smoke",
+  "device": "cuda",
+  "best_checkpoint": "/workspace/VLAarchtests/artifacts/outputs/interaction/rlbench_open_drawer_interaction_smoke/checkpoint_best.pt",
+  "final_train_total": 0.1628243774175644,
+  "final_val_total": 0.05217807739973068,
+  "train_dataset": {
+    "dataset_root": "/workspace/VLAarchtests/artifacts/data/rlbench_smoke_open_drawer",
+    "tasks": [
+      "open_drawer"
+    ],
+    "episode_indices": [
+      0
+    ],
+    "num_episodes": 1,
+    "num_samples": 1,
+    "resolution": 128,
+    "chunk_size": 4,
+    "proprio_dim": 32,
+    "history_steps": 6
+  },
+  "val_dataset": {
+    "dataset_root": "/workspace/VLAarchtests/artifacts/data/rlbench_smoke_open_drawer",
+    "tasks": [
+      "open_drawer"
+    ],
+    "episode_indices": [
+      1
+    ],
+    "num_episodes": 1,
+    "num_samples": 1,
+    "resolution": 128,
+    "chunk_size": 4,
+    "proprio_dim": 32,
+    "history_steps": 6
+  },
+  "init_info": {
+    "path": "/workspace/VLAarchtests/artifacts/outputs/interaction/proxy_interaction_state_smoke/checkpoint_best.pt",
+    "missing_keys": [],
+    "unexpected_keys": []
+  },
+  "planner_mode": "proxy_pretrained",
+  "plan_enabled_for_train": false,
+  "plan_enabled_for_eval": true,
+  "frozen_modules": [
+    "interaction_head",
+    "world_model",
+    "planner"
+  ]
+}

artifacts/outputs/interaction/rlbench_open_drawer_rollout_eval/rollout_eval.json

@@ -0,0 +1,23 @@
+{
+  "checkpoint": "/workspace/VLAarchtests/artifacts/outputs/interaction/rlbench_open_drawer_interaction_smoke/checkpoint_best.pt",
+  "plan_requested": true,
+  "plan_applied": true,
+  "planner_mode": "proxy_pretrained",
+  "support_mode_conditioning": true,
+  "episodes_per_task": 1,
+  "episode_length": 2,
+  "resolution": 128,
+  "cameras": [
+    "front",
+    "wrist_left",
+    "wrist_right"
+  ],
+  "tasks": {
+    "open_drawer": {
+      "error": "A path could not be found because the target is outside of workspace.",
+      "mean_success": 0.0,
+      "mean_return": 0.0
+    }
+  },
+  "mean_success": 0.0
+}

artifacts/outputs/interaction/rlbench_open_drawer_rollout_eval/rollout_eval.md

@@ -0,0 +1,11 @@
+# RLBench Rollout Eval
+
+- Checkpoint: `/workspace/VLAarchtests/artifacts/outputs/interaction/rlbench_open_drawer_interaction_smoke/checkpoint_best.pt`
+- Plan requested: `True`
+- Plan applied: `True`
+- Support-mode conditioning: `True`
+- Mean success: `0.000`
+
+## Per-task
+
+- `open_drawer`: error=A path could not be found because the target is outside of workspace.

artifacts/outputs/interaction/smoke_checks/smoke_checks.json

@@ -0,0 +1,156 @@
+{
+  "proxy": {
+    "losses": {
+      "action": 0.2892940640449524,
+      "phase": 1.4684441089630127,
+      "arm_role": 1.4457536935806274,
+      "support_mode": 1.0726321935653687,
+      "corridor": 0.7346047163009644,
+      "persistence": 1.7679060697555542,
+      "disturbance": 0.1656816601753235,
+      "belief": 0.6104393005371094,
+      "reocclusion": 0.6754688024520874,
+      "uncertainty": 0.733498752117157,
+      "world_model": 3.4413514137268066,
+      "planner_success": 0.5609128475189209,
+      "planner_risk": 0.09320252388715744,
+      "planner_ranking": 0.06516511738300323,
+      "proposal_reconstruction": 0.36118850111961365,
+      "proposal_success": 0.7169663906097412,
+      "proposal_ranking": 0.1487310826778412,
+      "total": 1.251321792602539
+    },
+    "grad_norm": 1.7515980005264282,
+    "candidate_shape": [
+      2,
+      4,
+      4,
+      14
+    ],
+    "rollout_phase_shape": [
+      2,
+      4,
+      4,
+      5
+    ]
+  },
+  "rlbench": {
+    "losses": {
+      "action": 0.4834498167037964,
+      "world_model": 0.0,
+      "planner_success": 0.0,
+      "planner_risk": 0.0,
+      "planner_ranking": 0.0,
+      "proposal_reconstruction": 0.0,
+      "proposal_success": 0.0,
+      "proposal_ranking": 0.0,
+      "total": 0.4834498167037964
+    },
+    "grad_norm": 2.1583147048950195,
+    "candidate_shape": [
+      2,
+      4,
+      4,
+      14
+    ],
+    "rollout_phase_shape": [
+      2,
+      4,
+      4,
+      5
+    ],
+    "planner_enabled_for_eval": true,
+    "frozen_modules": [
+      "interaction_head",
+      "world_model",
+      "planner"
+    ]
+  },
+  "policy_config": {
+    "backbone": {
+      "model_name": "openai/clip-vit-base-patch32",
+      "hidden_dim": 64,
+      "max_text_tokens": 32,
+      "freeze_backbone": true,
+      "gradient_checkpointing": false,
+      "use_dummy_backbone": true
+    },
+    "fusion": {
+      "hidden_dim": 64,
+      "num_cameras": 3,
+      "num_layers": 2,
+      "num_heads": 4,
+      "ff_dim": 128,
+      "dropout": 0.1,
+      "proprio_dim": 32,
+      "proprio_tokens": 1
+    },
+    "memory": {
+      "hidden_dim": 64,
+      "history_steps": 6,
+      "num_layers": 2,
+      "dropout": 0.1,
+      "memory_bank_size": 4,
+      "num_heads": 4,
+      "max_history_steps": 8
+    },
+    "decoder": {
+      "hidden_dim": 64,
+      "num_heads": 4,
+      "num_layers": 2,
+      "ff_dim": 128,
+      "dropout": 0.1,
+      "chunk_size": 4,
+      "action_dim": 14,
+      "arm_action_dim": 7,
+      "num_candidates": 4,
+      "num_phases": 5,
+      "num_arm_roles": 4
+    },
+    "reveal_head": {
+      "hidden_dim": 64,
+      "num_support_modes": 3,
+      "num_approach_templates": 32,
+      "rollout_horizon": 3,
+      "belief_map_size": 32,
+      "field_size": 16,
+      "num_heads": 4,
+      "predict_belief_map": true,
+      "num_phases": 5,
+      "num_arm_roles": 4,
+      "num_interaction_tokens": 8
+    },
+    "world_model": {
+      "hidden_dim": 64,
+      "action_dim": 14,
+      "num_support_modes": 3,
+      "num_approach_templates": 32,
+      "rollout_horizon": 3,
+      "field_size": 16,
+      "num_heads": 4,
+      "num_phases": 5,
+      "num_arm_roles": 4,
+      "num_interaction_tokens": 8,
+      "belief_map_size": 32,
+      "predict_belief_map": true
+    },
+    "planner": {
+      "hidden_dim": 64,
+      "num_candidates": 4,
+      "action_dim": 14,
+      "num_support_modes": 3,
+      "utility_margin": 0.1,
+      "corridor_weight": 1.0,
+      "persistence_weight": 0.5,
+      "proposal_weight": 0.5,
+      "task_progress_weight": 0.75,
+      "disturbance_weight": 0.75,
+      "reocclusion_weight": 0.5,
+      "visibility_weight": 0.25,
+      "num_heads": 4,
+      "num_layers": 2,
+      "num_phases": 5,
+      "num_arm_roles": 4
+    }
+  }
+}

artifacts/outputs/interaction_rolefix/proxy_interaction_state_smoke/config_resolved.yaml

@@ -0,0 +1,120 @@
+experiment_name: proxy_interaction_state_smoke
+output_dir: /workspace/VLAarchtests/artifacts/outputs/interaction
+device: cuda
+seed: 13
+data:
+  proxies:
+  - foliage_proxy
+  - bag_proxy
+  - cloth_proxy
+  resolution: 64
+  train_episodes_per_proxy: 6
+  val_episodes_per_proxy: 2
+  train_dataset_path: /workspace/VLAarchtests/artifacts/data/reveal_proxy/proxy_train_smoke_v4.pt
+  val_dataset_path: /workspace/VLAarchtests/artifacts/data/reveal_proxy/proxy_val_smoke_v4.pt
+  rebuild_dataset: false
+  chunk_horizon: 4
+  rollout_horizon: 3
+  history_steps: 6
+  planner_candidates: 4
+  seed: 13
+optim:
+  epochs: 2
+  batch_size: 8
+  num_workers: 0
+  lr: 0.001
+  weight_decay: 0.0001
+trainer:
+  policy_type: interaction_state
+  use_bf16: true
+  grad_clip_norm: 1.0
+  freeze_backbone: true
+  gradient_checkpointing: false
+  plan_during_train: true
+  plan_during_eval: true
+  support_mode_conditioning: true
+  planner_mode: trainable
+policy:
+  backbone:
+    model_name: openai/clip-vit-base-patch32
+    hidden_dim: 64
+    max_text_tokens: 32
+    freeze_backbone: true
+    gradient_checkpointing: false
+    use_dummy_backbone: true
+  fusion:
+    hidden_dim: 64
+    num_cameras: 3
+    num_layers: 2
+    num_heads: 4
+    ff_dim: 128
+    dropout: 0.1
+    proprio_dim: 32
+    proprio_tokens: 1
+  memory:
+    hidden_dim: 64
+    history_steps: 6
+    num_layers: 2
+    dropout: 0.1
+    memory_bank_size: 4
+    num_heads: 4
+    max_history_steps: 8
+  decoder:
+    hidden_dim: 64
+    num_heads: 4
+    num_layers: 2
+    ff_dim: 128
+    dropout: 0.1
+    chunk_size: 4
+    action_dim: 14
+    arm_action_dim: 7
+    num_candidates: 4
+    num_phases: 5
+    num_arm_roles: 4
+  reveal_head:
+    hidden_dim: 64
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 3
+    belief_map_size: 32
+    field_size: 16
+    num_heads: 4
+    predict_belief_map: true
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+  world_model:
+    hidden_dim: 64
+    action_dim: 14
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 3
+    field_size: 16
+    num_heads: 4
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+  planner:
+    hidden_dim: 64
+    num_candidates: 4
+    action_dim: 14
+    num_support_modes: 3
+    utility_margin: 0.1
+    num_heads: 4
+    num_layers: 2
+    num_phases: 5
+    num_arm_roles: 4
+loss_weights:
+  action: 1.0
+  support_mode: 0.15
+  corridor: 0.2
+  persistence: 0.1
+  disturbance: 0.1
+  world_model: 0.25
+  belief: 0.05
+  planner_success: 0.2
+  planner_risk: 0.1
+  planner_ranking: 0.1
+  proposal_reconstruction: 0.2
+  proposal_success: 0.1
+  proposal_ranking: 0.1

artifacts/outputs/interaction_rolefix/proxy_interaction_state_smoke/diagnostics/proxy_diagnostics.json

@@ -0,0 +1,7 @@
+{
+  "planner_top1_accuracy": 0.5,
+  "planner_regret": 0.28291308879852295,
+  "risk_calibration_mse": 0.011351427994668484,
+  "role_collapse_rate": 0.0,
+  "num_samples": 16
+}

artifacts/outputs/interaction_rolefix/proxy_interaction_state_smoke/metrics.json

@@ -0,0 +1,88 @@
+[
+  {
+    "epoch": 0,
+    "train": {
+      "action": 0.19658433113779342,
+      "arm_role": 1.238295819078173,
+      "belief": 0.5369629051004138,
+      "corridor": 0.5293313392571041,
+      "disturbance": 0.09085389758859362,
+      "persistence": 1.7996907574789864,
+      "phase": 1.0361328125,
+      "planner_ranking": 0.1484200656414032,
+      "planner_risk": 0.03149750775524548,
+      "planner_success": 0.6290038738931928,
+      "proposal_ranking": 0.12105858219521386,
+      "proposal_reconstruction": 0.26322201426540104,
+      "proposal_success": 0.5991358501570565,
+      "reocclusion": 0.65535911491939,
+      "support_mode": 0.8046875,
+      "total": 1.8467402798788888,
+      "uncertainty": 0.5706954428127834,
+      "world_model": 2.483790295464652
+    },
+    "val": {
+      "action": 0.07703214511275291,
+      "arm_role": 0.7396750450134277,
+      "belief": 0.391730472445488,
+      "corridor": 0.39731982350349426,
+      "disturbance": 0.0602701585739851,
+      "persistence": 1.7171077132225037,
+      "phase": 0.8182927072048187,
+      "planner_ranking": 0.10551176220178604,
+      "planner_risk": 0.010366585105657578,
+      "planner_success": 0.6895906329154968,
+      "proposal_ranking": 0.10800172761082649,
+      "proposal_reconstruction": 0.11226572096347809,
+      "proposal_success": 0.6531190276145935,
+      "reocclusion": 0.7202109694480896,
+      "support_mode": 0.8539398014545441,
+      "total": 1.366153061389923,
+      "uncertainty": 0.47223055362701416,
+      "world_model": 1.7127767205238342
+    }
+  },
+  {
+    "epoch": 1,
+    "train": {
+      "action": 0.0750974959560803,
+      "arm_role": 0.5170274334294456,
+      "belief": 0.26841118931770325,
+      "corridor": 0.31833365133830477,
+      "disturbance": 0.03495508113077709,
+      "persistence": 1.7808540718896049,
+      "phase": 0.825927734375,
+      "planner_ranking": 0.1236130531345095,
+      "planner_risk": 0.008478787228731173,
+      "planner_success": 0.6077252456120082,
+      "proposal_ranking": 0.08608398373637881,
+      "proposal_reconstruction": 0.111362986266613,
+      "proposal_success": 0.5995349713734218,
+      "reocclusion": 0.705062244619642,
+      "support_mode": 0.7960728236607143,
+      "total": 1.205997347831726,
+      "uncertainty": 0.387926242181233,
+      "world_model": 1.4581984792436873
+    },
+    "val": {
+      "action": 0.057233477011322975,
+      "arm_role": 0.269483357667923,
+      "belief": 0.14032869786024094,
+      "corridor": 0.2390739619731903,
+      "disturbance": 0.007302357582375407,
+      "persistence": 1.7315670251846313,
+      "phase": 0.692381888628006,
+      "planner_ranking": 0.10375471040606499,
+      "planner_risk": 0.011348946020007133,
+      "planner_success": 0.651288777589798,
+      "proposal_ranking": 0.1232626661658287,
+      "proposal_reconstruction": 0.1019212193787098,
+      "proposal_success": 0.6862722337245941,
+      "reocclusion": 0.7099874913692474,
+      "support_mode": 0.7302495241165161,
+      "total": 1.130429446697235,
+      "uncertainty": 0.2572660893201828,
+      "world_model": 1.5461857914924622
+    }
+  }
+]

artifacts/outputs/interaction_rolefix/reveal_eval_interaction_smoke/reveal_benchmark.json

@@ -0,0 +1,15 @@
+{
+  "interaction": {
+    "per_task_success": {
+      "foliage_proxy": 0.0,
+      "bag_proxy": 0.0,
+      "cloth_proxy": 0.0
+    },
+    "mean_success": 0.0,
+    "visibility_integral": 62.513198375701904,
+    "corridor_availability": 0.9871399253606796,
+    "reocclusion_rate": 0.0,
+    "persistence_horizon_mae": 2.8390470795333385,
+    "disturbance_cost": 0.9011474947134653
+  }
+}

artifacts/outputs/interaction_rolefix/reveal_eval_interaction_smoke/reveal_benchmark.md

@@ -0,0 +1,13 @@
+# Reveal Proxy Benchmark
+
+## interaction
+- checkpoint: /workspace/VLAarchtests/artifacts/outputs/interaction_rolefix/proxy_interaction_state_smoke/checkpoint_best.pt
+- mean_success: 0.000
+- visibility_integral: 62.513
+- corridor_availability: 0.987
+- reocclusion_rate: 0.000
+- persistence_horizon_mae: 2.839
+- disturbance_cost: 0.901
+- foliage_proxy_success: 0.000
+- bag_proxy_success: 0.000
+- cloth_proxy_success: 0.000

artifacts/outputs/interaction_rolefix/smoke_checks/smoke_checks.json

@@ -0,0 +1,156 @@
+{
+  "proxy": {
+    "losses": {
+      "action": 0.6122138500213623,
+      "phase": 1.8211359977722168,
+      "arm_role": 1.5296992063522339,
+      "support_mode": 1.1458361148834229,
+      "corridor": 0.7023701071739197,
+      "persistence": 2.00616717338562,
+      "disturbance": 0.16314803063869476,
+      "belief": 0.789874792098999,
+      "reocclusion": 0.6530417203903198,
+      "uncertainty": 0.7662769556045532,
+      "world_model": 3.8060429096221924,
+      "planner_success": 0.8667193651199341,
+      "planner_risk": 0.040573738515377045,
+      "planner_ranking": 0.26484477519989014,
+      "proposal_reconstruction": 0.7444740533828735,
+      "proposal_success": 0.8012435436248779,
+      "proposal_ranking": 0.07552901655435562,
+      "total": 1.9635474681854248
+    },
+    "grad_norm": 3.2980921268463135,
+    "candidate_shape": [
+      2,
+      4,
+      4,
+      14
+    ],
+    "rollout_phase_shape": [
+      2,
+      4,
+      4,
+      5
+    ]
+  },
+  "rlbench": {
+    "losses": {
+      "action": 0.44427579641342163,
+      "world_model": 0.0,
+      "planner_success": 0.0,
+      "planner_risk": 0.0,
+      "planner_ranking": 0.0,
+      "proposal_reconstruction": 0.0,
+      "proposal_success": 0.0,
+      "proposal_ranking": 0.0,
+      "total": 0.44427579641342163
+    },
+    "grad_norm": 2.2887489795684814,
+    "candidate_shape": [
+      2,
+      4,
+      4,
+      14
+    ],
+    "rollout_phase_shape": [
+      2,
+      4,
+      4,
+      5
+    ],
+    "planner_enabled_for_eval": true,
+    "frozen_modules": [
+      "interaction_head",
+      "world_model",
+      "planner"
+    ]
+  },
+  "policy_config": {
+    "backbone": {
+      "model_name": "openai/clip-vit-base-patch32",
+      "hidden_dim": 64,
+      "max_text_tokens": 32,
+      "freeze_backbone": true,
+      "gradient_checkpointing": false,
+      "use_dummy_backbone": true
+    },
+    "fusion": {
+      "hidden_dim": 64,
+      "num_cameras": 3,
+      "num_layers": 2,
+      "num_heads": 4,
+      "ff_dim": 128,
+      "dropout": 0.1,
+      "proprio_dim": 32,
+      "proprio_tokens": 1
+    },
+    "memory": {
+      "hidden_dim": 64,
+      "history_steps": 6,
+      "num_layers": 2,
+      "dropout": 0.1,
+      "memory_bank_size": 4,
+      "num_heads": 4,
+      "max_history_steps": 8
+    },
+    "decoder": {
+      "hidden_dim": 64,
+      "num_heads": 4,
+      "num_layers": 2,
+      "ff_dim": 128,
+      "dropout": 0.1,
+      "chunk_size": 4,
+      "action_dim": 14,
+      "arm_action_dim": 7,
+      "num_candidates": 4,
+      "num_phases": 5,
+      "num_arm_roles": 4
+    },
+    "reveal_head": {
+      "hidden_dim": 64,
+      "num_support_modes": 3,
+      "num_approach_templates": 32,
+      "rollout_horizon": 3,
+      "belief_map_size": 32,
+      "field_size": 16,
+      "num_heads": 4,
+      "predict_belief_map": true,
+      "num_phases": 5,
+      "num_arm_roles": 4,
+      "num_interaction_tokens": 8
+    },
+    "world_model": {
+      "hidden_dim": 64,
+      "action_dim": 14,
+      "num_support_modes": 3,
+      "num_approach_templates": 32,
+      "rollout_horizon": 3,
+      "field_size": 16,
+      "num_heads": 4,
+      "num_phases": 5,
+      "num_arm_roles": 4,
+      "num_interaction_tokens": 8,
+      "belief_map_size": 32,
+      "predict_belief_map": true
+    },
+    "planner": {
+      "hidden_dim": 64,
+      "num_candidates": 4,
+      "action_dim": 14,
+      "num_support_modes": 3,
+      "utility_margin": 0.1,
+      "corridor_weight": 1.0,
+      "persistence_weight": 0.5,
+      "proposal_weight": 0.5,
+      "task_progress_weight": 0.75,
+      "disturbance_weight": 0.75,
+      "reocclusion_weight": 0.5,
+      "visibility_weight": 0.25,
+      "num_heads": 4,
+      "num_layers": 2,
+      "num_phases": 5,
+      "num_arm_roles": 4
+    }
+  }
+}

artifacts/outputs/interaction_rolefix_full/proxy_interaction_state/config_resolved.yaml

@@ -0,0 +1,120 @@
+experiment_name: proxy_interaction_state
+output_dir: /workspace/VLAarchtests/artifacts/outputs/interaction
+device: cuda
+seed: 13
+data:
+  proxies:
+  - foliage_proxy
+  - bag_proxy
+  - cloth_proxy
+  resolution: 96
+  train_episodes_per_proxy: 48
+  val_episodes_per_proxy: 16
+  train_dataset_path: /workspace/VLAarchtests/artifacts/data/reveal_proxy/proxy_train_v4_noleak_counterfactual.pt
+  val_dataset_path: /workspace/VLAarchtests/artifacts/data/reveal_proxy/proxy_val_v4_noleak_counterfactual.pt
+  rebuild_dataset: false
+  chunk_horizon: 8
+  rollout_horizon: 5
+  history_steps: 6
+  planner_candidates: 8
+  seed: 13
+optim:
+  epochs: 8
+  batch_size: 16
+  num_workers: 0
+  lr: 0.001
+  weight_decay: 0.0001
+trainer:
+  policy_type: interaction_state
+  use_bf16: true
+  grad_clip_norm: 1.0
+  freeze_backbone: true
+  gradient_checkpointing: false
+  plan_during_train: true
+  plan_during_eval: true
+  support_mode_conditioning: true
+  planner_mode: trainable
+policy:
+  backbone:
+    model_name: openai/clip-vit-base-patch32
+    hidden_dim: 128
+    max_text_tokens: 32
+    freeze_backbone: true
+    gradient_checkpointing: false
+    use_dummy_backbone: true
+  fusion:
+    hidden_dim: 128
+    num_cameras: 3
+    num_layers: 2
+    num_heads: 4
+    ff_dim: 256
+    dropout: 0.1
+    proprio_dim: 32
+    proprio_tokens: 1
+  memory:
+    hidden_dim: 128
+    history_steps: 6
+    num_layers: 2
+    dropout: 0.1
+    memory_bank_size: 4
+    num_heads: 4
+    max_history_steps: 8
+  decoder:
+    hidden_dim: 128
+    num_heads: 4
+    num_layers: 2
+    ff_dim: 256
+    dropout: 0.1
+    chunk_size: 8
+    action_dim: 14
+    arm_action_dim: 7
+    num_candidates: 8
+    num_phases: 5
+    num_arm_roles: 4
+  reveal_head:
+    hidden_dim: 128
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 5
+    belief_map_size: 32
+    field_size: 16
+    num_heads: 4
+    predict_belief_map: true
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+  world_model:
+    hidden_dim: 128
+    action_dim: 14
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 5
+    field_size: 16
+    num_heads: 4
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+  planner:
+    hidden_dim: 128
+    num_candidates: 8
+    action_dim: 14
+    num_support_modes: 3
+    utility_margin: 0.1
+    num_heads: 4
+    num_layers: 2
+    num_phases: 5
+    num_arm_roles: 4
+loss_weights:
+  action: 1.0
+  support_mode: 0.15
+  corridor: 0.2
+  persistence: 0.1
+  disturbance: 0.1
+  world_model: 0.25
+  belief: 0.05
+  planner_success: 0.2
+  planner_risk: 0.1
+  planner_ranking: 0.1
+  proposal_reconstruction: 0.2
+  proposal_success: 0.1
+  proposal_ranking: 0.1

artifacts/outputs/interaction_rolefix_full/proxy_interaction_state/diagnostics/proxy_diagnostics.json

@@ -0,0 +1,7 @@
+{
+  "planner_top1_accuracy": 0.29365079365079366,
+  "planner_regret": 0.08918909728527069,
+  "risk_calibration_mse": 0.009334742091596127,
+  "role_collapse_rate": 0.0,
+  "num_samples": 126
+}

artifacts/outputs/interaction_rolefix_full/proxy_interaction_state/metrics.json

@@ -0,0 +1,346 @@
+[
+  {
+    "epoch": 0,
+    "train": {
+      "action": 0.10215476644225419,
+      "arm_role": 0.15410413238290252,
+      "belief": 0.2275839550420642,
+      "corridor": 0.31619268904129666,
+      "disturbance": 0.03557364640194768,
+      "persistence": 4.14529608686765,
+      "phase": 0.853579930961132,
+      "planner_ranking": 0.1402236126984159,
+      "planner_risk": 0.014521602657623589,
+      "planner_success": 0.6516827717423439,
+      "proposal_ranking": 0.12396896040687959,
+      "proposal_reconstruction": 0.15117586574827632,
+      "proposal_success": 0.6129913156231245,
+      "reocclusion": 0.6800185019771258,
+      "support_mode": 0.750041052699089,
+      "total": 1.7794813563426335,
+      "uncertainty": 0.2749920679877202,
+      "world_model": 2.930093298355738
+    },
+    "val": {
+      "action": 0.024464906891807914,
+      "arm_role": 0.00021468327213369776,
+      "belief": 0.11063466407358646,
+      "corridor": 0.24291332066059113,
+      "disturbance": 0.0021771492902189493,
+      "persistence": 4.035085469484329,
+      "phase": 0.7842688634991646,
+      "planner_ranking": 0.09380591381341219,
+      "planner_risk": 0.010528087033890188,
+      "planner_success": 0.6487544104456902,
+      "proposal_ranking": 0.09357322473078966,
+      "proposal_reconstruction": 0.05812322115525603,
+      "proposal_success": 0.6223367899656296,
+      "reocclusion": 0.6689130887389183,
+      "support_mode": 0.6824429258704185,
+      "total": 1.57107612490654,
+      "uncertainty": 0.04649291839450598,
+      "world_model": 2.8396184891462326
+    }
+  },
+  {
+    "epoch": 1,
+    "train": {
+      "action": 0.026911177905276418,
+      "arm_role": 9.164462486902873e-05,
+      "belief": 0.1127748250340422,
+      "corridor": 0.2482043243944645,
+      "disturbance": 0.0027894085894028344,
+      "persistence": 4.123300572236379,
+      "phase": 0.6966945106784502,
+      "planner_ranking": 0.1299859213953217,
+      "planner_risk": 0.011287257443958273,
+      "planner_success": 0.5469006697336832,
+      "proposal_ranking": 0.09497337446858485,
+      "proposal_reconstruction": 0.0546847361450394,
+      "proposal_success": 0.6040425598621368,
+      "reocclusion": 0.6738451148072878,
+      "support_mode": 0.6803697869181633,
+      "total": 1.4895438154538472,
+      "uncertainty": 0.04838159323359529,
+      "world_model": 2.5581150303284326
+    },
+    "val": {
+      "action": 0.0219610589556396,
+      "arm_role": 4.7797779643588e-05,
+      "belief": 0.11248616222292185,
+      "corridor": 0.24107512645423412,
+      "disturbance": 0.00229803760885261,
+      "persistence": 3.881433367729187,
+      "phase": 0.6858366876840591,
+      "planner_ranking": 0.10271583311259747,
+      "planner_risk": 0.010726989770773798,
+      "planner_success": 0.5016077607870102,
+      "proposal_ranking": 0.10926759615540504,
+      "proposal_reconstruction": 0.04846362583339214,
+      "proposal_success": 0.6263654828071594,
+      "reocclusion": 0.6623825207352638,
+      "support_mode": 0.6905809789896011,
+      "total": 1.4332728832960129,
+      "uncertainty": 0.03836950287222862,
+      "world_model": 2.490630403161049
+    }
+  },
+  {
+    "epoch": 2,
+    "train": {
+      "action": 0.028564381878823042,
+      "arm_role": 4.20472629230062e-05,
+      "belief": 0.11768585878113906,
+      "corridor": 0.2521046095838149,
+      "disturbance": 0.004006393282907084,
+      "persistence": 4.002537985642751,
+      "phase": 0.6817111050089201,
+      "planner_ranking": 0.14222958280394474,
+      "planner_risk": 0.010941210687936595,
+      "planner_success": 0.51358712464571,
+      "proposal_ranking": 0.10021264447520177,
+      "proposal_reconstruction": 0.05586679264282187,
+      "proposal_success": 0.6042261434098085,
+      "reocclusion": 0.6711050619681677,
+      "support_mode": 0.6735592807332674,
+      "total": 1.4198974122603734,
+      "uncertainty": 0.04468160099349916,
+      "world_model": 2.342899337410927
+    },
+    "val": {
+      "action": 0.03662527119740844,
+      "arm_role": 3.9326611840806436e-05,
+      "belief": 0.11038805078715086,
+      "corridor": 0.23578420095145702,
+      "disturbance": 0.0031197248463286087,
+      "persistence": 3.7882876694202423,
+      "phase": 0.677653007209301,
+      "planner_ranking": 0.12569222133606672,
+      "planner_risk": 0.010821192641742527,
+      "planner_success": 0.5561830475926399,
+      "proposal_ranking": 0.09295124560594559,
+      "proposal_reconstruction": 0.06929249316453934,
+      "proposal_success": 0.6254506856203079,
+      "reocclusion": 0.6604316681623459,
+      "support_mode": 0.6677745059132576,
+      "total": 1.2636573165655136,
+      "uncertainty": 0.02768313675187528,
+      "world_model": 1.7483449429273605
+    }
+  },
+  {
+    "epoch": 3,
+    "train": {
+      "action": 0.030749366773913305,
+      "arm_role": 4.2617077118241774e-05,
+      "belief": 0.11198224344601233,
+      "corridor": 0.24353070246676603,
+      "disturbance": 0.004747987113660201,
+      "persistence": 3.7688617209593454,
+      "phase": 0.6659654900431633,
+      "planner_ranking": 0.12002094245205323,
+      "planner_risk": 0.010106622396657864,
+      "planner_success": 0.472127099831899,
+      "proposal_ranking": 0.09478530536095302,
+      "proposal_reconstruction": 0.059262806084007025,
+      "proposal_success": 0.6104264035820961,
+      "reocclusion": 0.6585924625396729,
+      "support_mode": 0.6661049971977869,
+      "total": 1.3116108278433483,
+      "uncertainty": 0.02409021647569413,
+      "world_model": 2.0509919623533883
+    },
+    "val": {
+      "action": 0.020845535211265087,
+      "arm_role": 4.311565771786263e-05,
+      "belief": 0.105616076849401,
+      "corridor": 0.23376570269465446,
+      "disturbance": 0.002147116305422969,
+      "persistence": 3.7260473370552063,
+      "phase": 0.6661699637770653,
+      "planner_ranking": 0.10419025458395481,
+      "planner_risk": 0.010339341242797673,
+      "planner_success": 0.46252304688096046,
+      "proposal_ranking": 0.09330006875097752,
+      "proposal_reconstruction": 0.04469671146944165,
+      "proposal_success": 0.6214711964130402,
+      "reocclusion": 0.6415831819176674,
+      "support_mode": 0.6682936698198318,
+      "total": 1.2371356934309006,
+      "uncertainty": 0.021665064617991447,
+      "world_model": 1.8418128192424774
+    }
+  },
+  {
+    "epoch": 4,
+    "train": {
+      "action": 0.023547588226695854,
+      "arm_role": 4.4845532405209575e-05,
+      "belief": 0.10733066033571959,
+      "corridor": 0.2384033997853597,
+      "disturbance": 0.0032167407577314102,
+      "persistence": 3.8173932135105133,
+      "phase": 0.6745133176445961,
+      "planner_ranking": 0.10369156207889318,
+      "planner_risk": 0.01009264924020196,
+      "planner_success": 0.4860173078874747,
+      "proposal_ranking": 0.09702414429436128,
+      "proposal_reconstruction": 0.0478425626643002,
+      "proposal_success": 0.6053199768066406,
+      "reocclusion": 0.6491402039925257,
+      "support_mode": 0.6703156977891922,
+      "total": 1.2156481097141902,
+      "uncertainty": 0.024805960594676435,
+      "world_model": 1.684378559390704
+    },
+    "val": {
+      "action": 0.02007221127860248,
+      "arm_role": 6.0773097175115254e-05,
+      "belief": 0.10199161060154438,
+      "corridor": 0.22850653901696205,
+      "disturbance": 0.0017478264489909634,
+      "persistence": 3.6291932463645935,
+      "phase": 0.6642125844955444,
+      "planner_ranking": 0.09935609716922045,
+      "planner_risk": 0.009536529192700982,
+      "planner_success": 0.45760077610611916,
+      "proposal_ranking": 0.09855884779244661,
+      "proposal_reconstruction": 0.04397509805858135,
+      "proposal_success": 0.6215623244643211,
+      "reocclusion": 0.6404022425413132,
+      "support_mode": 0.6641640961170197,
+      "total": 1.153723917901516,
+      "uncertainty": 0.02464893041178584,
+      "world_model": 1.56255304813385
+    }
+  },
+  {
+    "epoch": 5,
+    "train": {
+      "action": 0.022455799548576277,
+      "arm_role": 5.3321499156785045e-05,
+      "belief": 0.10393211245536804,
+      "corridor": 0.23205090935031572,
+      "disturbance": 0.0021382822984984764,
+      "persistence": 3.760234226783117,
+      "phase": 0.6825721388061842,
+      "planner_ranking": 0.11921679352720578,
+      "planner_risk": 0.010239185998216271,
+      "planner_success": 0.467387559513251,
+      "proposal_ranking": 0.09636436806370814,
+      "proposal_reconstruction": 0.0463158117296795,
+      "proposal_success": 0.604705015818278,
+      "reocclusion": 0.6451127852002779,
+      "support_mode": 0.6723414833347002,
+      "total": 1.1677857091029484,
+      "uncertainty": 0.017458321526646614,
+      "world_model": 1.5345012545585632
+    },
+    "val": {
+      "action": 0.020456953439861536,
+      "arm_role": 5.5397130381606985e-05,
+      "belief": 0.100255505181849,
+      "corridor": 0.22343307547271252,
+      "disturbance": 0.0018057649213005789,
+      "persistence": 3.819442391395569,
+      "phase": 0.6870930641889572,
+      "planner_ranking": 0.12271129060536623,
+      "planner_risk": 0.009507962386123836,
+      "planner_success": 0.47849390283226967,
+      "proposal_ranking": 0.09397839661687613,
+      "proposal_reconstruction": 0.04395680082961917,
+      "proposal_success": 0.6256562769412994,
+      "reocclusion": 0.6624282449483871,
+      "support_mode": 0.6657112762331963,
+      "total": 1.3062576353549957,
+      "uncertainty": 0.007776510377880186,
+      "world_model": 2.0670085549354553
+    }
+  },
+  {
+    "epoch": 6,
+    "train": {
+      "action": 0.022894175878415506,
+      "arm_role": 5.4112812904349994e-05,
+      "belief": 0.10060124254475038,
+      "corridor": 0.23040568207701048,
+      "disturbance": 0.0020979453014054648,
+      "persistence": 3.749631315469742,
+      "phase": 0.6835000167290369,
+      "planner_ranking": 0.10540719423443079,
+      "planner_risk": 0.011304946111825606,
+      "planner_success": 0.43665214255452156,
+      "proposal_ranking": 0.09107906681795915,
+      "proposal_reconstruction": 0.04669830979158481,
+      "proposal_success": 0.6054539605975151,
+      "reocclusion": 0.6455749049782753,
+      "support_mode": 0.6700228725870451,
+      "total": 1.1786824663480122,
+      "uncertainty": 0.014146532688755542,
+      "world_model": 1.6150691360235214
+    },
+    "val": {
+      "action": 0.02070643357001245,
+      "arm_role": 4.2248018871759996e-05,
+      "belief": 0.10063061211258173,
+      "corridor": 0.22452155128121376,
+      "disturbance": 0.0018244763487018645,
+      "persistence": 3.6125852465629578,
+      "phase": 0.6734225451946259,
+      "planner_ranking": 0.0999936256557703,
+      "planner_risk": 0.010059075022581965,
+      "planner_success": 0.45402833819389343,
+      "proposal_ranking": 0.0882191276177764,
+      "proposal_reconstruction": 0.04391673672944307,
+      "proposal_success": 0.624024972319603,
+      "reocclusion": 0.6362400725483894,
+      "support_mode": 0.664807952940464,
+      "total": 1.134026937186718,
+      "uncertainty": 0.009506796195637435,
+      "world_model": 1.4956171810626984
+    }
+  },
+  {
+    "epoch": 7,
+    "train": {
+      "action": 0.02265403970765571,
+      "arm_role": 4.537651936213175e-05,
+      "belief": 0.10131731970856588,
+      "corridor": 0.22949850497146448,
+      "disturbance": 0.0021897068994197375,
+      "persistence": 3.6896196703116098,
+      "phase": 0.6795123865207037,
+      "planner_ranking": 0.09418578383823235,
+      "planner_risk": 0.00923546745131413,
+      "planner_success": 0.42756689339876175,
+      "proposal_ranking": 0.09173912275582552,
+      "proposal_reconstruction": 0.04657370550557971,
+      "proposal_success": 0.6038760592540106,
+      "reocclusion": 0.6364429369568825,
+      "support_mode": 0.6676999057332674,
+      "total": 1.1160989701747894,
+      "uncertainty": 0.01297911632961283,
+      "world_model": 1.4062732805808384
+    },
+    "val": {
+      "action": 0.023696284973993897,
+      "arm_role": 3.5256778573966585e-05,
+      "belief": 0.09744055476039648,
+      "corridor": 0.22240561619400978,
+      "disturbance": 0.0017758552203304134,
+      "persistence": 3.608371615409851,
+      "phase": 0.6682962626218796,
+      "planner_ranking": 0.10720822960138321,
+      "planner_risk": 0.009278025652747601,
+      "planner_success": 0.4377236105501652,
+      "proposal_ranking": 0.09098756965249777,
+      "proposal_reconstruction": 0.04804757330566645,
+      "proposal_success": 0.6270129010081291,
+      "reocclusion": 0.6341496035456657,
+      "support_mode": 0.6644193530082703,
+      "total": 1.0928071364760399,
+      "uncertainty": 0.015584498294629157,
+      "world_model": 1.3288657069206238
+    }
+  }
+]

artifacts/outputs/interaction_rolefix_full/reveal_eval_interaction/reveal_benchmark.json

@@ -0,0 +1,15 @@
+{
+  "interaction": {
+    "per_task_success": {
+      "foliage_proxy": 0.0,
+      "bag_proxy": 0.0,
+      "cloth_proxy": 0.0
+    },
+    "mean_success": 0.0,
+    "visibility_integral": 64.04203844070435,
+    "corridor_availability": 0.9837105721235275,
+    "reocclusion_rate": 0.0,
+    "persistence_horizon_mae": 1.7914747282500483,
+    "disturbance_cost": 0.8335009200705422
+  }
+}

artifacts/outputs/interaction_rolefix_full/reveal_eval_interaction/reveal_benchmark.md

@@ -0,0 +1,13 @@
+# Reveal Proxy Benchmark
+
+## interaction
+- checkpoint: /workspace/VLAarchtests/artifacts/outputs/interaction_rolefix_full/proxy_interaction_state/checkpoint_best.pt
+- mean_success: 0.000
+- visibility_integral: 64.042
+- corridor_availability: 0.984
+- reocclusion_rate: 0.000
+- persistence_horizon_mae: 1.791
+- disturbance_cost: 0.834
+- foliage_proxy_success: 0.000
+- bag_proxy_success: 0.000
+- cloth_proxy_success: 0.000