Add files using upload-large-folder tool

.gitattributes

@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+code/VLAarchtests2_code/VLAarchtests/code/reveal_vla_bimanual/eval/__pycache__/run_maniskill_pickclutter_smoke.cpython-311.pyc filter=lfs diff=lfs merge=lfs -text
+code/VLAarchtests2_code/VLAarchtests/code/reveal_vla_bimanual/eval/__pycache__/run_maniskill_bridge_retrieval_smoke.cpython-311.pyc filter=lfs diff=lfs merge=lfs -text
+reports/proxy_rank_only_live/active/reveal_benchmark.json filter=lfs diff=lfs merge=lfs -text

code/VLAarchtests2_code/VLAarchtests/code/reveal_vla_bimanual/eval/__pycache__/run_maniskill_bridge_retrieval_smoke.cpython-311.pyc

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a85cd23beb92a6be6a17a67a984917ed0a03bb54761fd9a853554bfd4aa0860
+size 130350

code/VLAarchtests2_code/VLAarchtests/code/reveal_vla_bimanual/eval/__pycache__/run_maniskill_pickclutter_smoke.cpython-311.pyc

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b4da5c78421ec938985ec16690dce1c0ae0a8f65da6097478e09059c1556f58
+size 131928

data/maniskill_bridge_retrieval/bag_bridge_smoke_v1/episode_splits.json

@@ -0,0 +1,98 @@
+{
+  "eval": [
+    172000,
+    172001,
+    172002,
+    172003,
+    172004,
+    172005,
+    172006,
+    172007,
+    172008,
+    172009,
+    172010,
+    172011,
+    172012,
+    172013,
+    172014,
+    172015,
+    172016,
+    172017,
+    172018,
+    172019,
+    172020,
+    172021,
+    172022,
+    172023,
+    172024,
+    172025,
+    172026,
+    172027,
+    172028,
+    172029,
+    172030,
+    172031,
+    172032,
+    172033,
+    172034,
+    172035,
+    172036,
+    172037,
+    172038,
+    172039,
+    172040,
+    172041,
+    172042,
+    172043,
+    172044,
+    172045,
+    172046,
+    172047,
+    172048,
+    172049
+  ],
+  "train": [
+    170000,
+    170001,
+    170002,
+    170003,
+    170004,
+    170005,
+    170006,
+    170007,
+    170008,
+    170009,
+    170010,
+    170011,
+    170012,
+    170013,
+    170014,
+    170015,
+    170016,
+    170017,
+    170018,
+    170019,
+    170020,
+    170021,
+    170022,
+    170023,
+    170024,
+    170025,
+    170026,
+    170027,
+    170028,
+    170029,
+    170030,
+    170031
+  ],
+  "val": [
+    171000,
+    171001,
+    171002,
+    171003,
+    171004,
+    171005,
+    171006,
+    171007
+  ]
+}

data/maniskill_bridge_retrieval/bag_bridge_smoke_v1/train.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64c854519a672f3190914069d9100442facc32789aa6ecca7147315b216a1aba
+size 907932312

data/maniskill_bridge_retrieval/bag_bridge_smoke_v1/val.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d97a77b8b583d1d13907ffa67f052ccaa9460b4d79026fc688a3ed24f3112e3
+size 329139344

data/maniskill_bridge_retrieval/cloth_bridge_smoke_v1/episode_splits.json

@@ -0,0 +1,98 @@
+{
+  "eval": [
+    170104,
+    170105,
+    170106,
+    170107,
+    170109,
+    170113,
+    170117,
+    170121,
+    170124,
+    170125,
+    170129,
+    170131,
+    170137,
+    170138,
+    170139,
+    170140,
+    170145,
+    170150,
+    170151,
+    170153,
+    170154,
+    170155,
+    170166,
+    170171,
+    170173,
+    170175,
+    170177,
+    170180,
+    170181,
+    170184,
+    170185,
+    170186,
+    170198,
+    170200,
+    170205,
+    170207,
+    170208,
+    170212,
+    170214,
+    170217,
+    170223,
+    170226,
+    170228,
+    170229,
+    170234,
+    170237,
+    170239,
+    170241,
+    170242,
+    170243
+  ],
+  "train": [
+    170000,
+    170005,
+    170006,
+    170008,
+    170012,
+    170016,
+    170019,
+    170020,
+    170028,
+    170031,
+    170032,
+    170035,
+    170036,
+    170037,
+    170038,
+    170044,
+    170045,
+    170049,
+    170050,
+    170053,
+    170054,
+    170055,
+    170057,
+    170059,
+    170061,
+    170066,
+    170067,
+    170069,
+    170073,
+    170077,
+    170079,
+    170081
+  ],
+  "val": [
+    170089,
+    170091,
+    170092,
+    170095,
+    170099,
+    170100,
+    170102,
+    170103
+  ]
+}

data/maniskill_bridge_retrieval/cloth_bridge_smoke_v1/train.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:926f4def9d473a355a5f9dc669055a4a5656071ee0fe9d6647fb6b9c160f3b95
+size 1389837080

data/maniskill_bridge_retrieval/cloth_bridge_smoke_v1/val.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a69eb63dbf6f1ca19f69437634c4931b73e4923a1d644617a2179a28813822e8
+size 347516816

data/maniskill_pickclutter/_debug_one.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb55e2a1cfaf72250af4d8436ac7323a66aa4b9ebc5890414a9d7bbca3180fc7
+size 37494636

data/maniskill_pickclutter/smoke_v1/episode_splits.json

@@ -0,0 +1,98 @@
+{
+  "eval": [
+    172000,
+    172001,
+    172002,
+    172003,
+    172004,
+    172005,
+    172006,
+    172007,
+    172008,
+    172009,
+    172010,
+    172011,
+    172012,
+    172013,
+    172014,
+    172015,
+    172016,
+    172017,
+    172018,
+    172019,
+    172020,
+    172021,
+    172022,
+    172023,
+    172024,
+    172025,
+    172026,
+    172027,
+    172028,
+    172029,
+    172030,
+    172031,
+    172032,
+    172033,
+    172034,
+    172035,
+    172036,
+    172037,
+    172038,
+    172039,
+    172040,
+    172041,
+    172042,
+    172043,
+    172044,
+    172045,
+    172046,
+    172047,
+    172048,
+    172049
+  ],
+  "train": [
+    170000,
+    170001,
+    170002,
+    170003,
+    170004,
+    170005,
+    170006,
+    170007,
+    170008,
+    170009,
+    170010,
+    170011,
+    170012,
+    170013,
+    170014,
+    170015,
+    170016,
+    170017,
+    170018,
+    170019,
+    170020,
+    170021,
+    170022,
+    170023,
+    170024,
+    170025,
+    170026,
+    170027,
+    170028,
+    170029,
+    170030,
+    170031
+  ],
+  "val": [
+    171000,
+    171001,
+    171002,
+    171003,
+    171004,
+    171005,
+    171006,
+    171007
+  ]
+}

data/maniskill_pickclutter/smoke_v1/train.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eecd3d3acfc05c6bb1aa54fd0b2a6c9771103aa6cf8453face688b99a56c2f3c
+size 1197054040

data/maniskill_pickclutter/smoke_v1/val.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba290e8ba892a41518ff56852db3804b4d85de6aaf5962ef857428e906b993b1
+size 341454800

data/maniskill_pickclutter/smoke_v2/episode_splits.json

@@ -0,0 +1,98 @@
+{
+  "eval": [
+    172000,
+    172001,
+    172002,
+    172003,
+    172004,
+    172005,
+    172006,
+    172007,
+    172008,
+    172009,
+    172010,
+    172011,
+    172012,
+    172013,
+    172014,
+    172015,
+    172016,
+    172017,
+    172018,
+    172019,
+    172020,
+    172021,
+    172022,
+    172023,
+    172024,
+    172025,
+    172026,
+    172027,
+    172028,
+    172029,
+    172030,
+    172031,
+    172032,
+    172033,
+    172034,
+    172035,
+    172036,
+    172037,
+    172038,
+    172039,
+    172040,
+    172041,
+    172042,
+    172043,
+    172044,
+    172045,
+    172046,
+    172047,
+    172048,
+    172049
+  ],
+  "train": [
+    170000,
+    170001,
+    170002,
+    170003,
+    170004,
+    170005,
+    170006,
+    170007,
+    170008,
+    170009,
+    170010,
+    170011,
+    170012,
+    170013,
+    170014,
+    170015,
+    170016,
+    170017,
+    170018,
+    170019,
+    170020,
+    170021,
+    170022,
+    170023,
+    170024,
+    170025,
+    170026,
+    170027,
+    170028,
+    170029,
+    170030,
+    170031
+  ],
+  "val": [
+    171000,
+    171001,
+    171002,
+    171003,
+    171004,
+    171005,
+    171006,
+    171007
+  ]
+}

data/maniskill_pickclutter/smoke_v2/train.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a484fda7cb224d1a1bada8b32cc7c0971cb9f8339e89fa3942c01d2f180d2b0
+size 1197206872

data/maniskill_pickclutter/smoke_v2/val.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6973f7ea8a1aafe7b01dc933226ab8d82f28bee2e2dec2d167f402a2c17347a
+size 341498704

data/maniskill_pickclutter/smoke_v3/episode_splits.json

@@ -0,0 +1,98 @@
+{
+  "eval": [
+    172000,
+    172001,
+    172002,
+    172003,
+    172004,
+    172005,
+    172006,
+    172007,
+    172008,
+    172009,
+    172010,
+    172011,
+    172012,
+    172013,
+    172014,
+    172015,
+    172016,
+    172017,
+    172018,
+    172019,
+    172020,
+    172021,
+    172022,
+    172023,
+    172024,
+    172025,
+    172026,
+    172027,
+    172028,
+    172029,
+    172030,
+    172031,
+    172032,
+    172033,
+    172034,
+    172035,
+    172036,
+    172037,
+    172038,
+    172039,
+    172040,
+    172041,
+    172042,
+    172043,
+    172044,
+    172045,
+    172046,
+    172047,
+    172048,
+    172049
+  ],
+  "train": [
+    170000,
+    170001,
+    170002,
+    170003,
+    170004,
+    170005,
+    170006,
+    170007,
+    170008,
+    170009,
+    170010,
+    170011,
+    170012,
+    170013,
+    170014,
+    170015,
+    170016,
+    170017,
+    170018,
+    170019,
+    170020,
+    170021,
+    170022,
+    170023,
+    170024,
+    170025,
+    170026,
+    170027,
+    170028,
+    170029,
+    170030,
+    170031
+  ],
+  "val": [
+    171000,
+    171001,
+    171002,
+    171003,
+    171004,
+    171005,
+    171006,
+    171007
+  ]
+}

data/maniskill_pickclutter/smoke_v3/train.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ecff2fb107f703fdaf0662bc5913b9c4e8326cd8cd92ee24f9a73a946247188
+size 1201790040

data/maniskill_pickclutter/smoke_v3/val.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bf0859f51001b2f4a152a9f00509d8ecb7bd486ec97619704b1844cfab2db6d
+size 342798032

data/reveal_proxy/proxy_train_clip224_v6_rgbd_stage3_phase_fast.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d195345f2dc6dfcdf666a4dca99ad717670cd4cb24b816c54b892137c5dd6906
+size 140638392

data/reveal_proxy/proxy_train_clip224_v6_rgbd_stage3_phase_fast_transition.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5d10aca6b616e13c713dddfa70e1a378e340caf13e2a825dc2a5e23b394b38e
+size 240587364

data/reveal_proxy/proxy_train_clip224_v6_rgbd_stage3_phase_rebuild128_seed17.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a14d47b0e5c2766739b9236258f10ffee24638fec98389ce1290be788840132f
+size 2682625452

data/reveal_proxy/proxy_val_clip224_v6_rgbd_stage3_phase_fast.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe8e7c4d10e4afe8f3cf211288159b07f4120b6c88fbfd6ef74f29ca8ba9eb76
+size 46282928

data/reveal_proxy/proxy_val_clip224_v6_rgbd_stage3_phase_fast_transition.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d90e7247fa635a47164b76df5a496c93307aa39b04183d294b90b2ed7e870c2
+size 79014044

data/reveal_proxy/proxy_val_clip224_v6_rgbd_stage3_phase_rebuild128_seed17.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:752e0111dd5427257a97ad8f7264b389fd8414d6d77cbf513a87b0cf7acd22e9
+size 675590116

outputs/adapter_proxy/proxy_adapter_wrapped_clip_base_fast_seed17/checkpoint_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:377c2ccd2d15193fcaf5203661e7353fc69bd48084e3564f245e496579a0c841
+size 878963364

outputs/adapter_proxy/proxy_adapter_wrapped_clip_base_fast_seed17/config_resolved.yaml

@@ -0,0 +1,173 @@
+experiment_name: proxy_adapter_wrapped_clip_base_fast_seed17
+output_dir: /workspace/workspace/outputs/adapter_proxy
+device: cuda
+seed: 17
+init_checkpoint: /workspace/workspace/VLAarchtests2/VLAarchtests/artifacts/outputs/r3d_handoff_phase/proxy_interaction_r3d_stage3_clip_rgbd_handoff_compact_phase_seed17/checkpoint_best.pt
+init_strict: false
+data:
+  proxies:
+  - foliage_proxy
+  - bag_proxy
+  - cloth_proxy
+  resolution: 224
+  dataset_version: reveal_proxy_v6_rgbd_elastic_state_phase_fast
+  train_episodes_per_proxy: 12
+  val_episodes_per_proxy: 4
+  train_dataset_path: /workspace/workspace/data/reveal_proxy/proxy_train_clip224_v6_rgbd_stage3_phase_fast.pt
+  val_dataset_path: /workspace/workspace/data/reveal_proxy/proxy_val_clip224_v6_rgbd_stage3_phase_fast.pt
+  rebuild_dataset: false
+  chunk_horizon: 8
+  rollout_horizon: 5
+  history_steps: 6
+  planner_candidates: 8
+  seed: 17
+optim:
+  epochs: 2
+  batch_size: 4
+  num_workers: 8
+  lr: 0.0001
+  weight_decay: 0.0001
+trainer:
+  policy_type: adapter_wrapped
+  training_regime: adapter_train_frozen_trunk
+  eval_mode: adapter_active
+  adapter_mode: adapter_active
+  adapter_use_transition_model: false
+  adapter_use_task_conditioning: true
+  use_bf16: true
+  grad_clip_norm: 1.0
+  freeze_backbone: true
+  gradient_checkpointing: false
+  plan_during_train: false
+  plan_during_eval: false
+  support_mode_conditioning: true
+  planner_mode: false
+  use_depth: true
+  use_world_model: false
+  use_role_tokens: true
+  compute_equivariance_probe: false
+  trainable_parameter_prefixes:
+  - adapter.state_head
+  - adapter.proposal_prior
+  - adapter.planner
+policy:
+  backbone:
+    model_name: openai/clip-vit-base-patch32
+    hidden_dim: 512
+    max_text_tokens: 32
+    freeze_backbone: true
+    gradient_checkpointing: false
+    use_dummy_backbone: false
+  fusion:
+    hidden_dim: 512
+    num_cameras: 3
+    num_layers: 4
+    num_heads: 8
+    ff_dim: 2048
+    dropout: 0.1
+    proprio_dim: 32
+    proprio_tokens: 1
+  memory:
+    hidden_dim: 512
+    action_dim: 14
+    history_steps: 6
+    scene_history_steps: 3
+    belief_history_steps: 8
+    num_layers: 2
+    dropout: 0.1
+    memory_bank_size: 4
+    scene_bank_size: 2
+    belief_bank_size: 2
+    num_heads: 8
+    max_history_steps: 8
+    reveal_cache_steps: 4
+    reveal_cache_decay: 0.7
+  decoder:
+    hidden_dim: 512
+    num_heads: 8
+    num_layers: 4
+    ff_dim: 2048
+    dropout: 0.1
+    chunk_size: 8
+    action_dim: 14
+    arm_action_dim: 7
+    num_candidates: 8
+    num_phases: 5
+    num_arm_roles: 4
+    num_proposal_modes: 7
+    planner_top_k: 4
+    proposal_delta_scale: 0.2
+    proposal_slot_scale: 0.05
+  reveal_head:
+    hidden_dim: 512
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 5
+    belief_map_size: 32
+    field_size: 16
+    num_heads: 8
+    predict_belief_map: true
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+    num_tasks: 4
+  world_model:
+    hidden_dim: 512
+    action_dim: 14
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 5
+    field_size: 16
+    num_heads: 8
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+    belief_map_size: 32
+    predict_belief_map: true
+    scene_bank_size: 2
+    belief_bank_size: 2
+    rollout_mode: compact_rollout
+    num_tasks: 4
+    lightweight_field_size: 4
+  planner:
+    hidden_dim: 512
+    num_candidates: 8
+    action_dim: 14
+    num_support_modes: 3
+    utility_margin: 0.1
+    num_heads: 8
+    num_layers: 2
+    num_phases: 5
+    num_arm_roles: 4
+    top_k: 4
+    adapter_confidence_threshold: 0.45
+loss_weights:
+  action: 1.0
+  phase: 0.08
+  arm_role: 0.08
+  support_mode: 0.08
+  corridor: 0.12
+  persistence: 0.06
+  disturbance: 0.06
+  world_model: 0.0
+  transition: 0.0
+  belief: 0.05
+  visibility: 0.05
+  clearance: 0.06
+  support_stability: 0.06
+  reocclusion: 0.06
+  occluder_contact: 0.05
+  grasp_affordance: 0.05
+  planner_success: 0.15
+  planner_risk: 0.08
+  planner_ranking: 0.15
+  proposal_reconstruction: 0.08
+  proposal_success: 0.1
+  proposal_ranking: 0.12
+  proposal_mode: 0.08
+  proposal_diversity: 0.05
+  role_swap_consistency: 0.0
+  task_metrics: 0.06
+  gate: 0.05
+  distillation: 0.05
+  calibration: 0.02

outputs/adapter_proxy/proxy_adapter_wrapped_clip_base_fast_seed17/metrics.json

@@ -0,0 +1,140 @@
+[
+  {
+    "epoch": 0,
+    "train": {
+      "action": 1.1780137238295183,
+      "arm_role": 0.000544056080402895,
+      "belief": 0.10274084074341733,
+      "calibration": 0.0,
+      "clearance": 0.08112246429790622,
+      "corridor": 0.21243907782532598,
+      "distillation": 0.0036539296447501883,
+      "disturbance": 0.0010930091615908009,
+      "gate": 0.0,
+      "grasp_affordance": 0.011060374242294094,
+      "occluder_contact": 0.19354943348013837,
+      "persistence": 0.29602919886415097,
+      "phase": 0.1456924275211666,
+      "planner_ranking": 1.1046701566032742,
+      "planner_risk": 0.03252584584381269,
+      "planner_success": 0.5002943964108176,
+      "proposal_diversity": 0.0,
+      "proposal_mode": 0.9053098727827487,
+      "proposal_ranking": 0.7633599224297897,
+      "proposal_reconstruction": 1.1813416908616605,
+      "proposal_success": 0.5018493273983831,
+      "reocclusion": 0.1370238650428212,
+      "role_swap_consistency": 0.0,
+      "support_mode": 0.0010332910049170175,
+      "support_stability": 0.13264792088581168,
+      "task_metrics": 0.07693366929078879,
+      "total": 1.8312026676924333,
+      "transition": 0.0,
+      "uncertainty": 1.4312560102039045e-05,
+      "visibility": 0.096126823645571,
+      "world_model": 0.0
+    },
+    "val": {
+      "action": 1.146972581744194,
+      "arm_role": 2.7849786739864157e-05,
+      "belief": 0.09928969945758581,
+      "calibration": 0.0,
+      "clearance": 0.07546275667846203,
+      "corridor": 0.18693614657968283,
+      "distillation": 0.005982774979202077,
+      "disturbance": 0.0012652746545427362,
+      "gate": 0.0,
+      "grasp_affordance": 0.009092151012737304,
+      "occluder_contact": 0.19199086539447308,
+      "persistence": 0.4173499735770747,
+      "phase": 0.20510842488147318,
+      "planner_ranking": 1.0746948570013046,
+      "planner_risk": 0.03205434698611498,
+      "planner_success": 0.3765582703053951,
+      "proposal_diversity": 0.0,
+      "proposal_mode": 0.5553285405039787,
+      "proposal_ranking": 0.6613346468657255,
+      "proposal_reconstruction": 1.1140409670770168,
+      "proposal_success": 0.32496484369039536,
+      "reocclusion": 0.2021030569449067,
+      "role_swap_consistency": 0.0,
+      "support_mode": 0.00011286496555840131,
+      "support_stability": 0.13265474420040846,
+      "task_metrics": 0.06524855340830982,
+      "total": 1.7250810116529465,
+      "transition": 0.0,
+      "uncertainty": 8.913456255754681e-06,
+      "visibility": 0.09269411116838455,
+      "world_model": 0.0
+    }
+  },
+  {
+    "epoch": 1,
+    "train": {
+      "action": 1.1840074995289678,
+      "arm_role": 1.7842088946844857e-05,
+      "belief": 0.10108890773161598,
+      "calibration": 0.0,
+      "clearance": 0.08066983359015506,
+      "corridor": 0.20431885726587928,
+      "distillation": 0.005328163808292668,
+      "disturbance": 0.000988402207440231,
+      "gate": 0.0,
+      "grasp_affordance": 0.010460576832132496,
+      "occluder_contact": 0.19120351322319196,
+      "persistence": 0.20984708754669712,
+      "phase": 0.1270662468412648,
+      "planner_ranking": 1.051699793857077,
+      "planner_risk": 0.03183994928131933,
+      "planner_success": 0.37528212303700653,
+      "proposal_diversity": 0.0,
+      "proposal_mode": 0.541168266016504,
+      "proposal_ranking": 0.7413897125617318,
+      "proposal_reconstruction": 1.1529877976230953,
+      "proposal_success": 0.273181245378826,
+      "reocclusion": 0.11955958685797194,
+      "role_swap_consistency": 0.0,
+      "support_mode": 0.00014792317929475203,
+      "support_stability": 0.1314481108084969,
+      "task_metrics": 0.07543641668946846,
+      "total": 1.744326695151951,
+      "transition": 0.0,
+      "uncertainty": 7.94198708297739e-06,
+      "visibility": 0.09458825672450273,
+      "world_model": 0.0
+    },
+    "val": {
+      "action": 1.1787440478801727,
+      "arm_role": 1.3783465302452669e-05,
+      "belief": 0.0974554605782032,
+      "calibration": 0.0,
+      "clearance": 0.0746708307415247,
+      "corridor": 0.18591812625527382,
+      "distillation": 0.0038922334788367152,
+      "disturbance": 0.0005819438138132682,
+      "gate": 0.0,
+      "grasp_affordance": 0.008575586834922433,
+      "occluder_contact": 0.19005733728408813,
+      "persistence": 0.4048172008187976,
+      "phase": 0.24421580568014178,
+      "planner_ranking": 1.0271672308444977,
+      "planner_risk": 0.03108011605218053,
+      "planner_success": 0.3713325075805187,
+      "proposal_diversity": 0.0,
+      "proposal_mode": 0.46797188371419907,
+      "proposal_ranking": 0.6800601556897163,
+      "proposal_reconstruction": 1.0902876928448677,
+      "proposal_success": 0.25984624214470387,
+      "reocclusion": 0.19258547481149435,
+      "role_swap_consistency": 0.0,
+      "support_mode": 0.00014510085156871355,
+      "support_stability": 0.13228781055659056,
+      "task_metrics": 0.06339579145424068,
+      "total": 1.7367750853300095,
+      "transition": 0.0,
+      "uncertainty": 6.649694360483238e-06,
+      "visibility": 0.09114759508520365,
+      "world_model": 0.0
+    }
+  }
+]

outputs/adapter_proxy/proxy_adapter_wrapped_clip_base_reuse128_seed17/checkpoint_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2b7436f5685a38203440d3c84fc7f86908ddba389b844866554b4d0c0fd0768
+size 878965604

outputs/adapter_proxy/proxy_adapter_wrapped_clip_base_reuse128_seed17/metrics.json

@@ -0,0 +1,278 @@
+[
+  {
+    "epoch": 0,
+    "train": {
+      "action": 1.1828932802216345,
+      "arm_role": 0.00244398444339226,
+      "belief": 0.10072019552232839,
+      "calibration": 0.0,
+      "clearance": 0.07946077994063121,
+      "corridor": 0.21543118382702356,
+      "distillation": 0.00042247207064432005,
+      "disturbance": 0.0009066167868626844,
+      "gate": 0.0,
+      "grasp_affordance": 0.011442071496031615,
+      "occluder_contact": 0.19184747789086415,
+      "persistence": 0.5456274578801724,
+      "phase": 0.1889389944928033,
+      "planner_ranking": 0.8968874569199666,
+      "planner_risk": 0.03290799349358603,
+      "planner_success": 0.35506935793311656,
+      "proposal_diversity": 0.0,
+      "proposal_mode": 0.7599493966383093,
+      "proposal_ranking": 1.4915186276956767,
+      "proposal_reconstruction": 1.0803285907296574,
+      "proposal_success": 0.3194384900461726,
+      "reocclusion": 0.1872198152817598,
+      "role_swap_consistency": 0.0,
+      "support_mode": 0.4244060135689102,
+      "support_stability": 0.13155287654459977,
+      "task_metrics": 0.07493724777292804,
+      "total": 2.751452175509028,
+      "transition": 4.318220460114359,
+      "uncertainty": 1.531094441807496e-05,
+      "visibility": 0.09642757938689545,
+      "world_model": 0.0
+    },
+    "val": {
+      "action": 1.1680383563041687,
+      "arm_role": 0.0025612511759391054,
+      "belief": 0.09879593178629875,
+      "calibration": 0.0,
+      "clearance": 0.07741740134855112,
+      "corridor": 0.20817755659421286,
+      "distillation": 0.0,
+      "disturbance": 0.0007382428300237128,
+      "gate": 0.0,
+      "grasp_affordance": 0.010511041525751353,
+      "occluder_contact": 0.19018630186716715,
+      "persistence": 0.4509886346757412,
+      "phase": 0.1597365932694326,
+      "planner_ranking": 0.22907628491520882,
+      "planner_risk": 0.02909238338470459,
+      "planner_success": 0.18200772007306418,
+      "proposal_diversity": 0.0,
+      "proposal_mode": 0.71118057568868,
+      "proposal_ranking": 1.4729209462801616,
+      "proposal_reconstruction": 1.015290528535843,
+      "proposal_success": 0.2791739940643311,
+      "reocclusion": 0.16477556849519412,
+      "role_swap_consistency": 0.0,
+      "support_mode": 0.5340653051932652,
+      "support_stability": 0.12872510105371476,
+      "task_metrics": 0.06174707182993491,
+      "total": 2.407643111546834,
+      "transition": 3.39704422156016,
+      "uncertainty": 7.099100287177862e-06,
+      "visibility": 0.09383414511879286,
+      "world_model": 0.0
+    }
+  },
+  {
+    "epoch": 1,
+    "train": {
+      "action": 1.187044749740793,
+      "arm_role": 0.001233981896833587,
+      "belief": 0.09885497215916128,
+      "calibration": 0.0,
+      "clearance": 0.07787450506281451,
+      "corridor": 0.21069503738349224,
+      "distillation": 0.0,
+      "disturbance": 0.0007993320816102586,
+      "gate": 0.0,
+      "grasp_affordance": 0.0100274878874922,
+      "occluder_contact": 0.19033558541486242,
+      "persistence": 0.508021433908148,
+      "phase": 0.19023076729739413,
+      "planner_ranking": 0.058458461105322636,
+      "planner_risk": 0.03440776518976488,
+      "planner_success": 0.1257152666627359,
+      "proposal_diversity": 0.0,
+      "proposal_mode": 0.7171601638072679,
+      "proposal_ranking": 1.499033512187605,
+      "proposal_reconstruction": 1.066634831809196,
+      "proposal_success": 0.3018947724534684,
+      "reocclusion": 0.16926059677821248,
+      "role_swap_consistency": 0.0,
+      "support_mode": 0.4455214215426886,
+      "support_stability": 0.13059799138362668,
+      "task_metrics": 0.07159904390573502,
+      "total": 2.4211200485710336,
+      "transition": 3.487839874099283,
+      "uncertainty": 3.770016950513401e-06,
+      "visibility": 0.09318254963189614,
+      "world_model": 0.0
+    },
+    "val": {
+      "action": 1.1680383563041687,
+      "arm_role": 0.001657356577925384,
+      "belief": 0.09766801769534747,
+      "calibration": 0.0,
+      "clearance": 0.07670599135259787,
+      "corridor": 0.20785387406746547,
+      "distillation": 0.0,
+      "disturbance": 0.0007254338066559285,
+      "gate": 0.0,
+      "grasp_affordance": 0.009808245363334816,
+      "occluder_contact": 0.18903621584177016,
+      "persistence": 0.43403610289096833,
+      "phase": 0.17749264603480697,
+      "planner_ranking": 0.00962653555907309,
+      "planner_risk": 0.02840747827043136,
+      "planner_success": 0.0469651294251283,
+      "proposal_diversity": 0.0,
+      "proposal_mode": 0.5958098510901133,
+      "proposal_ranking": 1.567319353421529,
+      "proposal_reconstruction": 1.0027365585168202,
+      "proposal_success": 0.3119396299123764,
+      "reocclusion": 0.14939573630690575,
+      "role_swap_consistency": 0.0,
+      "support_mode": 0.38477273682753244,
+      "support_stability": 0.12813995343943438,
+      "task_metrics": 0.05784295691798131,
+      "total": 2.3466440041859946,
+      "transition": 3.402106682459513,
+      "uncertainty": 3.2218885041383296e-06,
+      "visibility": 0.09148541142543157,
+      "world_model": 0.0
+    }
+  },
+  {
+    "epoch": 2,
+    "train": {
+      "action": 1.187824563819821,
+      "arm_role": 0.0017524876263963075,
+      "belief": 0.09850409833573494,
+      "calibration": 0.0,
+      "clearance": 0.07750590865602013,
+      "corridor": 0.21022135673576042,
+      "distillation": 0.0,
+      "disturbance": 0.0008020720826393432,
+      "gate": 0.0,
+      "grasp_affordance": 0.009951516582841883,
+      "occluder_contact": 0.190022504630209,
+      "persistence": 0.5073582559448331,
+      "phase": 0.17974354623339506,
+      "planner_ranking": 0.009596662447169549,
+      "planner_risk": 0.03246875642603185,
+      "planner_success": 0.06673186843698266,
+      "proposal_diversity": 0.0,
+      "proposal_mode": 0.7036348676481167,
+      "proposal_ranking": 1.4990194234527459,
+      "proposal_reconstruction": 1.0593123075340976,
+      "proposal_success": 0.30170050113141034,
+      "reocclusion": 0.1706294410807245,
+      "role_swap_consistency": 0.0,
+      "support_mode": 0.4435207678490326,
+      "support_stability": 0.12954452590030782,
+      "task_metrics": 0.07019141574679803,
+      "total": 2.3952997061384824,
+      "transition": 3.4510987426052573,
+      "uncertainty": 2.649417712834203e-06,
+      "visibility": 0.09213429119657068,
+      "world_model": 0.0
+    },
+    "val": {
+      "action": 1.1680383563041687,
+      "arm_role": 0.0005777989087315897,
+      "belief": 0.09620878870288531,
+      "calibration": 0.0,
+      "clearance": 0.07562205567955971,
+      "corridor": 0.2099471464753151,
+      "distillation": 0.0,
+      "disturbance": 0.0008037402614718304,
+      "gate": 0.0,
+      "grasp_affordance": 0.009381201630458236,
+      "occluder_contact": 0.18789172718922298,
+      "persistence": 0.44771519377827645,
+      "phase": 0.15351878677805264,
+      "planner_ranking": 0.005908836016897112,
+      "planner_risk": 0.029111843556165695,
+      "planner_success": 0.030371779979517063,
+      "proposal_diversity": 0.0,
+      "proposal_mode": 0.6608088513215383,
+      "proposal_ranking": 1.519856317838033,
+      "proposal_reconstruction": 0.9984971513350804,
+      "proposal_success": 0.2899133563041687,
+      "reocclusion": 0.15338999405503273,
+      "role_swap_consistency": 0.0,
+      "support_mode": 0.4591325432062149,
+      "support_stability": 0.12738436510165532,
+      "task_metrics": 0.05577167191853126,
+      "total": 2.3411471287409467,
+      "transition": 3.3808055957158407,
+      "uncertainty": 1.560352771671584e-06,
+      "visibility": 0.08981477295358976,
+      "world_model": 0.0
+    }
+  },
+  {
+    "epoch": 3,
+    "train": {
+      "action": 1.1873075451169695,
+      "arm_role": 0.0010167556069400005,
+      "belief": 0.09699463875604276,
+      "calibration": 0.0,
+      "clearance": 0.0765939431280649,
+      "corridor": 0.21000426350271,
+      "distillation": 0.0,
+      "disturbance": 0.0008205439020564561,
+      "gate": 0.0,
+      "grasp_affordance": 0.009616962144886996,
+      "occluder_contact": 0.1890684860844572,
+      "persistence": 0.5268036977802756,
+      "phase": 0.18212753434141143,
+      "planner_ranking": 0.007861482998857102,
+      "planner_risk": 0.0305439497837249,
+      "planner_success": 0.0545816100632944,
+      "proposal_diversity": 0.0,
+      "proposal_mode": 0.7096028443144149,
+      "proposal_ranking": 1.49962230790563,
+      "proposal_reconstruction": 1.0570516235688154,
+      "proposal_success": 0.3012468101096754,
+      "reocclusion": 0.16893144916085637,
+      "role_swap_consistency": 0.0,
+      "support_mode": 0.43846767214166016,
+      "support_stability": 0.12901192851865492,
+      "task_metrics": 0.0706772211500827,
+      "total": 2.383075835324135,
+      "transition": 3.399705786664947,
+      "uncertainty": 1.833678168140796e-06,
+      "visibility": 0.09043271063255663,
+      "world_model": 0.0
+    },
+    "val": {
+      "action": 1.1680383563041687,
+      "arm_role": 0.0008160963848543664,
+      "belief": 0.09533951580524444,
+      "calibration": 0.0,
+      "clearance": 0.07521944617231686,
+      "corridor": 0.2074363355835279,
+      "distillation": 0.0,
+      "disturbance": 0.0007471947777958121,
+      "gate": 0.0,
+      "grasp_affordance": 0.009425108910848697,
+      "occluder_contact": 0.187281297147274,
+      "persistence": 0.42866156020512186,
+      "phase": 0.13389708844115375,
+      "planner_ranking": 0.007386005097456897,
+      "planner_risk": 0.03013829297075669,
+      "planner_success": 0.027494619445254404,
+      "proposal_diversity": 0.0,
+      "proposal_mode": 0.7145659645398458,
+      "proposal_ranking": 1.4651208639144897,
+      "proposal_reconstruction": 0.99560972849528,
+      "proposal_success": 0.29622272253036497,
+      "reocclusion": 0.15021706620852152,
+      "role_swap_consistency": 0.0,
+      "support_mode": 0.3665752013524373,
+      "support_stability": 0.12691180408000946,
+      "task_metrics": 0.056707360843817396,
+      "total": 2.3298022985458373,
+      "transition": 3.3876041332880655,
+      "uncertainty": 1.581879031557302e-06,
+      "visibility": 0.08887151132027309,
+      "world_model": 0.0
+    }
+  }
+]

outputs/adapter_proxy/proxy_adapter_wrapped_clip_rank_only_fast_seed17/checkpoint_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f52bffd6b1228221303a52fa31678c02c0df4cf0493aae456340f2c0fd19e66d
+size 878963300

outputs/adapter_proxy/proxy_adapter_wrapped_clip_rank_only_rebuild128_seed17/checkpoint_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bfd34e337b63d17b4a09d4c5b47a2b02484ae20ebd88d0c43c15e90f2a53eea
+size 878965540

outputs/adapter_proxy/proxy_adapter_wrapped_clip_rank_only_rebuild128_seed17/config_resolved.yaml

@@ -0,0 +1,170 @@
+experiment_name: proxy_adapter_wrapped_clip_rank_only_rebuild128_seed17
+output_dir: /workspace/workspace/outputs/adapter_proxy
+device: cuda
+seed: 17
+init_checkpoint: /workspace/workspace/VLAarchtests2/VLAarchtests/artifacts/outputs/r3d_handoff_phase/proxy_interaction_r3d_stage3_clip_rgbd_handoff_compact_phase_seed17/checkpoint_best.pt
+init_strict: false
+data:
+  proxies:
+  - foliage_proxy
+  - bag_proxy
+  - cloth_proxy
+  resolution: 224
+  dataset_version: reveal_proxy_v6_rgbd_elastic_state_phase
+  train_episodes_per_proxy: 128
+  val_episodes_per_proxy: 32
+  train_dataset_path: /workspace/workspace/data/reveal_proxy/proxy_train_clip224_v6_rgbd_stage3_phase_rebuild128_seed17.pt
+  val_dataset_path: /workspace/workspace/data/reveal_proxy/proxy_val_clip224_v6_rgbd_stage3_phase_rebuild128_seed17.pt
+  rebuild_dataset: true
+  chunk_horizon: 8
+  rollout_horizon: 5
+  history_steps: 6
+  planner_candidates: 8
+  seed: 17
+optim:
+  epochs: 4
+  batch_size: 8
+  num_workers: 32
+  lr: 5.0e-05
+  weight_decay: 0.0001
+trainer:
+  policy_type: adapter_wrapped
+  training_regime: proxy_rank_only
+  eval_mode: adapter_active
+  adapter_mode: adapter_active
+  adapter_use_transition_model: false
+  adapter_use_task_conditioning: true
+  use_bf16: true
+  grad_clip_norm: 1.0
+  freeze_backbone: true
+  gradient_checkpointing: false
+  plan_during_train: false
+  plan_during_eval: false
+  support_mode_conditioning: true
+  planner_mode: false
+  use_depth: true
+  use_world_model: false
+  use_role_tokens: true
+  compute_equivariance_probe: false
+  trainable_parameter_prefixes:
+  - adapter.proposal_prior
+  - adapter.planner
+policy:
+  backbone:
+    model_name: openai/clip-vit-base-patch32
+    hidden_dim: 512
+    max_text_tokens: 32
+    freeze_backbone: true
+    gradient_checkpointing: false
+    use_dummy_backbone: false
+  fusion:
+    hidden_dim: 512
+    num_cameras: 3
+    num_layers: 4
+    num_heads: 8
+    ff_dim: 2048
+    dropout: 0.1
+    proprio_dim: 32
+    proprio_tokens: 1
+  memory:
+    hidden_dim: 512
+    action_dim: 14
+    history_steps: 6
+    scene_history_steps: 3
+    belief_history_steps: 8
+    num_layers: 2
+    dropout: 0.1
+    memory_bank_size: 4
+    scene_bank_size: 2
+    belief_bank_size: 2
+    num_heads: 8
+    max_history_steps: 8
+  decoder:
+    hidden_dim: 512
+    num_heads: 8
+    num_layers: 4
+    ff_dim: 2048
+    dropout: 0.1
+    chunk_size: 8
+    action_dim: 14
+    arm_action_dim: 7
+    num_candidates: 8
+    num_phases: 5
+    num_arm_roles: 4
+    num_proposal_modes: 7
+    planner_top_k: 4
+    proposal_delta_scale: 0.2
+    proposal_slot_scale: 0.05
+  reveal_head:
+    hidden_dim: 512
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 5
+    belief_map_size: 32
+    field_size: 16
+    num_heads: 8
+    predict_belief_map: true
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+    num_tasks: 4
+  world_model:
+    hidden_dim: 512
+    action_dim: 14
+    num_support_modes: 3
+    num_approach_templates: 32
+    rollout_horizon: 5
+    field_size: 16
+    num_heads: 8
+    num_phases: 5
+    num_arm_roles: 4
+    num_interaction_tokens: 8
+    belief_map_size: 32
+    predict_belief_map: true
+    scene_bank_size: 2
+    belief_bank_size: 2
+    rollout_mode: compact_rollout
+    num_tasks: 4
+    lightweight_field_size: 4
+  planner:
+    hidden_dim: 512
+    num_candidates: 8
+    action_dim: 14
+    num_support_modes: 3
+    utility_margin: 0.1
+    num_heads: 8
+    num_layers: 2
+    num_phases: 5
+    num_arm_roles: 4
+    top_k: 4
+    adapter_confidence_threshold: 0.55
+loss_weights:
+  action: 0.5
+  phase: 0.0
+  arm_role: 0.0
+  support_mode: 0.0
+  corridor: 0.0
+  persistence: 0.0
+  disturbance: 0.0
+  world_model: 0.0
+  transition: 0.0
+  belief: 0.0
+  visibility: 0.0
+  clearance: 0.0
+  support_stability: 0.0
+  reocclusion: 0.0
+  occluder_contact: 0.0
+  grasp_affordance: 0.0
+  planner_success: 0.0
+  planner_risk: 0.0
+  planner_ranking: 0.2
+  proposal_reconstruction: 0.0
+  proposal_success: 0.1
+  proposal_ranking: 0.2
+  proposal_mode: 0.1
+  proposal_diversity: 0.02
+  role_swap_consistency: 0.0
+  task_metrics: 0.0
+  gate: 0.0
+  distillation: 0.05
+  calibration: 0.0

outputs/adapter_proxy/proxy_adapter_wrapped_clip_transition_fast_seed17/checkpoint_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c5fd7c44df9c0406fe34b2d94e1d21594273cb51d21e6a15fc5c82486c95200
+size 878964132

outputs/anchor_adapter_wrapped_dual_push_seed17/checkpoint_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31ba52c98865e44202c6a6d41e290c7ba0914da8aea1090c7d2e2a2f9b987a77
+size 811648748

outputs/maniskill_bag_bridge_smoke_v1/adapter_active_ft_seed17/checkpoint_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:62e3a1fd0956525ecf82a3c43386645545dc39770062b82ea94a4e7e23d9a296
+size 878857572

outputs/maniskill_bag_bridge_smoke_v1/adapter_active_ft_seed23/checkpoint_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2cc79d0a38e6bdf98609a7bf9dd2f42be5f07ce442911879ce00c1815f31815
+size 878856868

outputs/maniskill_bag_bridge_smoke_v1/trunk_only_ft_seed17/checkpoint_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc20a31e9014f71b1e5475f959c3561cc2cbb8557e6a2cd21bc7a1502192b889
+size 817967468

outputs/maniskill_bag_bridge_smoke_v1/trunk_only_ft_seed17/summary.json

@@ -0,0 +1,1104 @@
+{
+  "task": "bag",
+  "variant": "trunk_only_ft",
+  "checkpoint_path": "/workspace/workspace/outputs/maniskill_bag_bridge_smoke_v1/trunk_only_ft_seed17/checkpoint_best.pt",
+  "init_info": {
+    "path": "/workspace/workspace/VLAarchtests2/VLAarchtests/artifacts/outputs/r3d_handoff_phase/proxy_interaction_r3d_stage3_clip_rgbd_handoff_compact_phase_seed17/checkpoint_best.pt",
+    "loaded_keys": 489,
+    "skipped_shape_mismatch_keys": [
+      "memory.scene_memory.position_embedding",
+      "memory.scene_memory.bank_queries",
+      "memory.scene_memory.sequence_encoder.layers.0.self_attn.in_proj_weight",
+      "memory.scene_memory.sequence_encoder.layers.0.self_attn.in_proj_bias",
+      "memory.scene_memory.sequence_encoder.layers.0.self_attn.out_proj.weight",
+      "memory.scene_memory.sequence_encoder.layers.0.self_attn.out_proj.bias",
+      "memory.scene_memory.sequence_encoder.layers.0.linear1.weight",
+      "memory.scene_memory.sequence_encoder.layers.0.linear1.bias",
+      "memory.scene_memory.sequence_encoder.layers.0.linear2.weight",
+      "memory.scene_memory.sequence_encoder.layers.0.linear2.bias",
+      "memory.scene_memory.sequence_encoder.layers.0.norm1.weight",
+      "memory.scene_memory.sequence_encoder.layers.0.norm1.bias",
+      "memory.scene_memory.sequence_encoder.layers.0.norm2.weight",
+      "memory.scene_memory.sequence_encoder.layers.0.norm2.bias",
+      "memory.scene_memory.bank_attention.in_proj_weight",
+      "memory.scene_memory.bank_attention.in_proj_bias",
+      "memory.scene_memory.bank_attention.out_proj.weight",
+      "memory.scene_memory.bank_attention.out_proj.bias",
+      "memory.scene_memory.action_proj.0.weight",
+      "memory.scene_memory.action_proj.0.bias",
+      "memory.scene_memory.action_proj.1.weight",
+      "memory.scene_memory.action_proj.1.bias",
+      "memory.scene_memory.write_gate.0.weight",
+      "memory.scene_memory.write_gate.0.bias",
+      "memory.scene_memory.write_gate.1.weight",
+      "memory.scene_memory.write_gate.1.bias",
+      "memory.scene_memory.write_gate.3.weight",
+      "memory.scene_memory.write_gate.3.bias",
+      "memory.scene_memory.token_proj.0.weight",
+      "memory.scene_memory.token_proj.0.bias",
+      "memory.scene_memory.token_proj.1.weight",
+      "memory.scene_memory.token_proj.1.bias",
+      "memory.belief_memory.position_embedding",
+      "memory.belief_memory.bank_queries",
+      "memory.belief_memory.sequence_encoder.layers.0.self_attn.in_proj_weight",
+      "memory.belief_memory.sequence_encoder.layers.0.self_attn.in_proj_bias",
+      "memory.belief_memory.sequence_encoder.layers.0.self_attn.out_proj.weight",
+      "memory.belief_memory.sequence_encoder.layers.0.self_attn.out_proj.bias",
+      "memory.belief_memory.sequence_encoder.layers.0.linear1.weight",
+      "memory.belief_memory.sequence_encoder.layers.0.linear1.bias",
+      "memory.belief_memory.sequence_encoder.layers.0.linear2.weight",
+      "memory.belief_memory.sequence_encoder.layers.0.linear2.bias",
+      "memory.belief_memory.sequence_encoder.layers.0.norm1.weight",
+      "memory.belief_memory.sequence_encoder.layers.0.norm1.bias",
+      "memory.belief_memory.sequence_encoder.layers.0.norm2.weight",
+      "memory.belief_memory.sequence_encoder.layers.0.norm2.bias",
+      "memory.belief_memory.bank_attention.in_proj_weight",
+      "memory.belief_memory.bank_attention.in_proj_bias",
+      "memory.belief_memory.bank_attention.out_proj.weight",
+      "memory.belief_memory.bank_attention.out_proj.bias",
+      "memory.belief_memory.action_proj.0.weight",
+      "memory.belief_memory.action_proj.0.bias",
+      "memory.belief_memory.action_proj.1.weight",
+      "memory.belief_memory.action_proj.1.bias",
+      "memory.belief_memory.write_gate.0.weight",
+      "memory.belief_memory.write_gate.0.bias",
+      "memory.belief_memory.write_gate.1.weight",
+      "memory.belief_memory.write_gate.1.bias",
+      "memory.belief_memory.write_gate.3.weight",
+      "memory.belief_memory.write_gate.3.bias",
+      "memory.belief_memory.token_proj.0.weight",
+      "memory.belief_memory.token_proj.0.bias",
+      "memory.belief_memory.token_proj.1.weight",
+      "memory.belief_memory.token_proj.1.bias",
+      "decoder.arm_decoder.layers.0.self_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.0.self_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.0.self_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.0.self_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.0.multihead_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.0.multihead_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.0.multihead_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.0.multihead_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.0.linear1.weight",
+      "decoder.arm_decoder.layers.0.linear1.bias",
+      "decoder.arm_decoder.layers.0.linear2.weight",
+      "decoder.arm_decoder.layers.0.linear2.bias",
+      "decoder.arm_decoder.layers.0.norm1.weight",
+      "decoder.arm_decoder.layers.0.norm1.bias",
+      "decoder.arm_decoder.layers.0.norm2.weight",
+      "decoder.arm_decoder.layers.0.norm2.bias",
+      "decoder.arm_decoder.layers.0.norm3.weight",
+      "decoder.arm_decoder.layers.0.norm3.bias",
+      "decoder.arm_decoder.layers.1.self_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.1.self_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.1.self_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.1.self_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.1.multihead_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.1.multihead_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.1.multihead_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.1.multihead_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.1.linear1.weight",
+      "decoder.arm_decoder.layers.1.linear1.bias",
+      "decoder.arm_decoder.layers.1.linear2.weight",
+      "decoder.arm_decoder.layers.1.linear2.bias",
+      "decoder.arm_decoder.layers.1.norm1.weight",
+      "decoder.arm_decoder.layers.1.norm1.bias",
+      "decoder.arm_decoder.layers.1.norm2.weight",
+      "decoder.arm_decoder.layers.1.norm2.bias",
+      "decoder.arm_decoder.layers.1.norm3.weight",
+      "decoder.arm_decoder.layers.1.norm3.bias",
+      "decoder.arm_decoder.layers.2.self_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.2.self_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.2.self_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.2.self_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.2.multihead_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.2.multihead_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.2.multihead_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.2.multihead_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.2.linear1.weight",
+      "decoder.arm_decoder.layers.2.linear1.bias",
+      "decoder.arm_decoder.layers.2.linear2.weight",
+      "decoder.arm_decoder.layers.2.linear2.bias",
+      "decoder.arm_decoder.layers.2.norm1.weight",
+      "decoder.arm_decoder.layers.2.norm1.bias",
+      "decoder.arm_decoder.layers.2.norm2.weight",
+      "decoder.arm_decoder.layers.2.norm2.bias",
+      "decoder.arm_decoder.layers.2.norm3.weight",
+      "decoder.arm_decoder.layers.2.norm3.bias",
+      "decoder.arm_decoder.layers.3.self_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.3.self_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.3.self_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.3.self_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.3.multihead_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.3.multihead_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.3.multihead_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.3.multihead_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.3.linear1.weight",
+      "decoder.arm_decoder.layers.3.linear1.bias",
+      "decoder.arm_decoder.layers.3.linear2.weight",
+      "decoder.arm_decoder.layers.3.linear2.bias",
+      "decoder.arm_decoder.layers.3.norm1.weight",
+      "decoder.arm_decoder.layers.3.norm1.bias",
+      "decoder.arm_decoder.layers.3.norm2.weight",
+      "decoder.arm_decoder.layers.3.norm2.bias",
+      "decoder.arm_decoder.layers.3.norm3.weight",
+      "decoder.arm_decoder.layers.3.norm3.bias",
+      "decoder.arm_identity.weight",
+      "decoder.task_embedding.weight",
+      "decoder.phase_adapter.weight",
+      "decoder.phase_adapter.bias",
+      "decoder.role_adapter.weight",
+      "decoder.role_adapter.bias",
+      "decoder.context_proj.0.weight",
+      "decoder.context_proj.0.bias",
+      "decoder.context_proj.1.weight",
+      "decoder.context_proj.1.bias",
+      "decoder.arm_head.0.weight",
+      "decoder.arm_head.0.bias",
+      "decoder.arm_head.1.weight",
+      "decoder.arm_head.1.bias",
+      "decoder.arm_mean.weight",
+      "decoder.arm_mean.bias",
+      "decoder.arm_log_std.weight",
+      "decoder.arm_log_std.bias",
+      "decoder.proposal_mode_head.0.weight",
+      "decoder.proposal_mode_head.0.bias",
+      "decoder.proposal_mode_head.1.weight",
+      "decoder.proposal_mode_head.1.bias",
+      "decoder.proposal_mode_head.3.weight",
+      "decoder.proposal_mode_head.3.bias",
+      "decoder.proposal_mode_embeddings.weight",
+      "decoder.proposal_slot_embeddings.weight",
+      "decoder.mode_residual_heads.0.0.weight",
+      "decoder.mode_residual_heads.0.0.bias",
+      "decoder.mode_residual_heads.0.1.weight",
+      "decoder.mode_residual_heads.0.1.bias",
+      "decoder.mode_residual_heads.0.3.weight",
+      "decoder.mode_residual_heads.0.3.bias",
+      "decoder.mode_residual_heads.1.0.weight",
+      "decoder.mode_residual_heads.1.0.bias",
+      "decoder.mode_residual_heads.1.1.weight",
+      "decoder.mode_residual_heads.1.1.bias",
+      "decoder.mode_residual_heads.1.3.weight",
+      "decoder.mode_residual_heads.1.3.bias",
+      "decoder.mode_residual_heads.2.0.weight",
+      "decoder.mode_residual_heads.2.0.bias",
+      "decoder.mode_residual_heads.2.1.weight",
+      "decoder.mode_residual_heads.2.1.bias",
+      "decoder.mode_residual_heads.2.3.weight",
+      "decoder.mode_residual_heads.2.3.bias",
+      "decoder.mode_residual_heads.3.0.weight",
+      "decoder.mode_residual_heads.3.0.bias",
+      "decoder.mode_residual_heads.3.1.weight",
+      "decoder.mode_residual_heads.3.1.bias",
+      "decoder.mode_residual_heads.3.3.weight",
+      "decoder.mode_residual_heads.3.3.bias",
+      "decoder.mode_residual_heads.4.0.weight",
+      "decoder.mode_residual_heads.4.0.bias",
+      "decoder.mode_residual_heads.4.1.weight",
+      "decoder.mode_residual_heads.4.1.bias",
+      "decoder.mode_residual_heads.4.3.weight",
+      "decoder.mode_residual_heads.4.3.bias",
+      "decoder.mode_residual_heads.5.0.weight",
+      "decoder.mode_residual_heads.5.0.bias",
+      "decoder.mode_residual_heads.5.1.weight",
+      "decoder.mode_residual_heads.5.1.bias",
+      "decoder.mode_residual_heads.5.3.weight",
+      "decoder.mode_residual_heads.5.3.bias",
+      "decoder.mode_residual_heads.6.0.weight",
+      "decoder.mode_residual_heads.6.0.bias",
+      "decoder.mode_residual_heads.6.1.weight",
+      "decoder.mode_residual_heads.6.1.bias",
+      "decoder.mode_residual_heads.6.3.weight",
+      "decoder.mode_residual_heads.6.3.bias",
+      "decoder.slot_delta.0.weight",
+      "decoder.slot_delta.0.bias",
+      "decoder.slot_delta.1.weight",
+      "decoder.slot_delta.1.bias",
+      "decoder.slot_delta.3.weight",
+      "decoder.slot_delta.3.bias",
+      "decoder.proposal_score.0.weight",
+      "decoder.proposal_score.0.bias",
+      "decoder.proposal_score.1.weight",
+      "decoder.proposal_score.1.bias",
+      "decoder.proposal_score.3.weight",
+      "decoder.proposal_score.3.bias",
+      "elastic_state_head.interaction_queries",
+      "elastic_state_head.interaction_attention.in_proj_weight",
+      "elastic_state_head.interaction_attention.in_proj_bias",
+      "elastic_state_head.interaction_attention.out_proj.weight",
+      "elastic_state_head.interaction_attention.out_proj.bias",
+      "elastic_state_head.interaction_mlp.0.weight",
+      "elastic_state_head.interaction_mlp.0.bias",
+      "elastic_state_head.interaction_mlp.1.weight",
+      "elastic_state_head.interaction_mlp.1.bias",
+      "elastic_state_head.interaction_mlp.3.weight",
+      "elastic_state_head.interaction_mlp.3.bias",
+      "elastic_state_head.decoder.field_queries",
+      "elastic_state_head.decoder.field_attention.in_proj_weight",
+      "elastic_state_head.decoder.field_attention.in_proj_bias",
+      "elastic_state_head.decoder.field_attention.out_proj.weight",
+      "elastic_state_head.decoder.field_attention.out_proj.bias",
+      "elastic_state_head.decoder.field_mlp.0.weight",
+      "elastic_state_head.decoder.field_mlp.0.bias",
+      "elastic_state_head.decoder.field_mlp.1.weight",
+      "elastic_state_head.decoder.field_mlp.1.bias",
+      "elastic_state_head.decoder.field_mlp.3.weight",
+      "elastic_state_head.decoder.field_mlp.3.bias",
+      "elastic_state_head.decoder.summary_proj.0.weight",
+      "elastic_state_head.decoder.summary_proj.0.bias",
+      "elastic_state_head.decoder.summary_proj.1.weight",
+      "elastic_state_head.decoder.summary_proj.1.bias",
+      "elastic_state_head.decoder.phase_head.0.weight",
+      "elastic_state_head.decoder.phase_head.0.bias",
+      "elastic_state_head.decoder.phase_head.1.weight",
+      "elastic_state_head.decoder.phase_head.1.bias",
+      "elastic_state_head.decoder.phase_head.3.weight",
+      "elastic_state_head.decoder.phase_head.3.bias",
+      "elastic_state_head.decoder.arm_role_head.0.weight",
+      "elastic_state_head.decoder.arm_role_head.0.bias",
+      "elastic_state_head.decoder.arm_role_head.1.weight",
+      "elastic_state_head.decoder.arm_role_head.1.bias",
+      "elastic_state_head.decoder.arm_role_head.3.weight",
+      "elastic_state_head.decoder.arm_role_head.3.bias",
+      "elastic_state_head.decoder.arm_identity.weight",
+      "elastic_state_head.decoder.support_mode.0.weight",
+      "elastic_state_head.decoder.support_mode.0.bias",
+      "elastic_state_head.decoder.support_mode.1.weight",
+      "elastic_state_head.decoder.support_mode.1.bias",
+      "elastic_state_head.decoder.support_mode.3.weight",
+      "elastic_state_head.decoder.support_mode.3.bias",
+      "elastic_state_head.decoder.access_field.weight",
+      "elastic_state_head.decoder.access_field.bias",
+      "elastic_state_head.decoder.target_belief_field.weight",
+      "elastic_state_head.decoder.target_belief_field.bias",
+      "elastic_state_head.decoder.visibility_field.weight",
+      "elastic_state_head.decoder.visibility_field.bias",
+      "elastic_state_head.decoder.clearance_field.weight",
+      "elastic_state_head.decoder.clearance_field.bias",
+      "elastic_state_head.decoder.occluder_contact_field.weight",
+      "elastic_state_head.decoder.occluder_contact_field.bias",
+      "elastic_state_head.decoder.grasp_affordance_field.weight",
+      "elastic_state_head.decoder.grasp_affordance_field.bias",
+      "elastic_state_head.decoder.support_stability_field.weight",
+      "elastic_state_head.decoder.support_stability_field.bias",
+      "elastic_state_head.decoder.persistence_field.weight",
+      "elastic_state_head.decoder.persistence_field.bias",
+      "elastic_state_head.decoder.reocclusion_field.weight",
+      "elastic_state_head.decoder.reocclusion_field.bias",
+      "elastic_state_head.decoder.disturbance_field.weight",
+      "elastic_state_head.decoder.disturbance_field.bias",
+      "elastic_state_head.decoder.uncertainty_field.weight",
+      "elastic_state_head.decoder.uncertainty_field.bias",
+      "elastic_state_head.decoder.reocclusion_head.0.weight",
+      "elastic_state_head.decoder.reocclusion_head.0.bias",
+      "elastic_state_head.decoder.reocclusion_head.1.weight",
+      "elastic_state_head.decoder.reocclusion_head.1.bias",
+      "elastic_state_head.decoder.reocclusion_head.3.weight",
+      "elastic_state_head.decoder.reocclusion_head.3.bias",
+      "elastic_state_head.decoder.task_embedding.weight",
+      "elastic_state_head.decoder.task_field_affine.weight",
+      "elastic_state_head.decoder.task_field_affine.bias",
+      "elastic_state_head.decoder.task_summary_adapter.0.weight",
+      "elastic_state_head.decoder.task_summary_adapter.0.bias",
+      "elastic_state_head.decoder.task_summary_adapter.1.weight",
+      "elastic_state_head.decoder.task_summary_adapter.1.bias",
+      "elastic_state_head.decoder.task_phase_head.weight",
+      "elastic_state_head.decoder.task_phase_head.bias",
+      "elastic_state_head.decoder.task_support_head.weight",
+      "elastic_state_head.decoder.task_support_head.bias",
+      "elastic_state_head.decoder.task_reocclusion_head.weight",
+      "elastic_state_head.decoder.task_reocclusion_head.bias",
+      "elastic_state_head.decoder.task_metric_head.0.weight",
+      "elastic_state_head.decoder.task_metric_head.0.bias",
+      "elastic_state_head.decoder.task_metric_head.1.weight",
+      "elastic_state_head.decoder.task_metric_head.1.bias",
+      "elastic_state_head.decoder.task_metric_head.3.weight",
+      "elastic_state_head.decoder.task_metric_head.3.bias",
+      "world_model.state_encoder.0.weight",
+      "world_model.state_encoder.0.bias",
+      "world_model.state_encoder.1.weight",
+      "world_model.state_encoder.1.bias",
+      "world_model.scene_memory_proj.0.weight",
+      "world_model.scene_memory_proj.0.bias",
+      "world_model.scene_memory_proj.1.weight",
+      "world_model.scene_memory_proj.1.bias",
+      "world_model.belief_memory_proj.0.weight",
+      "world_model.belief_memory_proj.0.bias",
+      "world_model.belief_memory_proj.1.weight",
+      "world_model.belief_memory_proj.1.bias",
+      "world_model.action_encoder.0.weight",
+      "world_model.action_encoder.0.bias",
+      "world_model.action_encoder.1.weight",
+      "world_model.action_encoder.1.bias",
+      "world_model.transition.weight_ih",
+      "world_model.transition.weight_hh",
+      "world_model.transition.bias_ih",
+      "world_model.transition.bias_hh",
+      "world_model.scene_memory_update.weight",
+      "world_model.scene_memory_update.bias",
+      "world_model.belief_memory_update.weight",
+      "world_model.belief_memory_update.bias",
+      "world_model.compact_decoder.weight",
+      "world_model.compact_decoder.bias",
+      "world_model.target_belief_head.weight",
+      "world_model.target_belief_head.bias",
+      "world_model.visibility_head.weight",
+      "world_model.visibility_head.bias",
+      "world_model.clearance_head.weight",
+      "world_model.clearance_head.bias",
+      "world_model.occluder_contact_head.weight",
+      "world_model.occluder_contact_head.bias",
+      "world_model.grasp_affordance_head.weight",
+      "world_model.grasp_affordance_head.bias",
+      "world_model.support_stability_head.weight",
+      "world_model.support_stability_head.bias",
+      "world_model.persistence_head.weight",
+      "world_model.persistence_head.bias",
+      "world_model.reocclusion_head.weight",
+      "world_model.reocclusion_head.bias",
+      "world_model.disturbance_head.weight",
+      "world_model.disturbance_head.bias",
+      "world_model.uncertainty_head.weight",
+      "world_model.uncertainty_head.bias",
+      "world_model.access_head.weight",
+      "world_model.access_head.bias",
+      "world_model.task_embedding.weight",
+      "world_model.spatial_field_encoder.0.weight",
+      "world_model.spatial_field_encoder.0.bias",
+      "world_model.spatial_field_encoder.2.weight",
+      "world_model.spatial_field_encoder.2.bias",
+      "world_model.spatial_context_proj.0.weight",
+      "world_model.spatial_context_proj.0.bias",
+      "world_model.spatial_context_proj.1.weight",
+      "world_model.spatial_context_proj.1.bias",
+      "world_model.spatial_gate_z.weight",
+      "world_model.spatial_gate_z.bias",
+      "world_model.spatial_gate_r.weight",
+      "world_model.spatial_gate_r.bias",
+      "world_model.spatial_candidate.weight",
+      "world_model.spatial_candidate.bias",
+      "world_model.spatial_summary_proj.0.weight",
+      "world_model.spatial_summary_proj.0.bias",
+      "world_model.spatial_summary_proj.1.weight",
+      "world_model.spatial_summary_proj.1.bias",
+      "world_model.spatial_phase_head.weight",
+      "world_model.spatial_phase_head.bias",
+      "world_model.spatial_support_mode_head.weight",
+      "world_model.spatial_support_mode_head.bias",
+      "world_model.spatial_arm_role_head.weight",
+      "world_model.spatial_arm_role_head.bias",
+      "world_model.spatial_reocclusion_head.weight",
+      "world_model.spatial_reocclusion_head.bias",
+      "world_model.spatial_target_belief_head.weight",
+      "world_model.spatial_target_belief_head.bias",
+      "world_model.spatial_visibility_head.weight",
+      "world_model.spatial_visibility_head.bias",
+      "world_model.spatial_clearance_head.weight",
+      "world_model.spatial_clearance_head.bias",
+      "world_model.spatial_occluder_contact_head.weight",
+      "world_model.spatial_occluder_contact_head.bias",
+      "world_model.spatial_grasp_affordance_head.weight",
+      "world_model.spatial_grasp_affordance_head.bias",
+      "world_model.spatial_support_stability_head.weight",
+      "world_model.spatial_support_stability_head.bias",
+      "world_model.spatial_persistence_head.weight",
+      "world_model.spatial_persistence_head.bias",
+      "world_model.spatial_reocclusion_field_head.weight",
+      "world_model.spatial_reocclusion_field_head.bias",
+      "world_model.spatial_disturbance_head.weight",
+      "world_model.spatial_disturbance_head.bias",
+      "world_model.spatial_uncertainty_head.weight",
+      "world_model.spatial_uncertainty_head.bias",
+      "world_model.spatial_access_head.weight",
+      "world_model.spatial_access_head.bias",
+      "planner.residual.trunk.0.weight",
+      "planner.residual.trunk.0.bias",
+      "planner.residual.trunk.1.weight",
+      "planner.residual.trunk.1.bias",
+      "planner.residual.trunk.3.weight",
+      "planner.residual.trunk.3.bias",
+      "planner.residual.success_head.weight",
+      "planner.residual.success_head.bias",
+      "planner.residual.risk_head.weight",
+      "planner.residual.risk_head.bias",
+      "planner.residual.residual_head.weight",
+      "planner.residual.residual_head.bias"
+    ],
+    "remapped_keys": {},
+    "missing_keys": [
+      "memory.gru.weight_ih_l0",
+      "memory.gru.weight_hh_l0",
+      "memory.gru.bias_ih_l0",
+      "memory.gru.bias_hh_l0",
+      "memory.gru.weight_ih_l1",
+      "memory.gru.weight_hh_l1",
+      "memory.gru.bias_ih_l1",
+      "memory.gru.bias_hh_l1",
+      "memory.token_proj.0.weight",
+      "memory.token_proj.0.bias",
+      "memory.token_proj.1.weight",
+      "memory.token_proj.1.bias",
+      "memory.action_proj.0.weight",
+      "memory.action_proj.0.bias",
+      "memory.action_proj.1.weight",
+      "memory.action_proj.1.bias",
+      "decoder.actor_role_bias",
+      "decoder.revealer_decoder.layers.0.self_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.0.self_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.0.self_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.0.self_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.0.multihead_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.0.multihead_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.0.multihead_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.0.multihead_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.0.linear1.weight",
+      "decoder.revealer_decoder.layers.0.linear1.bias",
+      "decoder.revealer_decoder.layers.0.linear2.weight",
+      "decoder.revealer_decoder.layers.0.linear2.bias",
+      "decoder.revealer_decoder.layers.0.norm1.weight",
+      "decoder.revealer_decoder.layers.0.norm1.bias",
+      "decoder.revealer_decoder.layers.0.norm2.weight",
+      "decoder.revealer_decoder.layers.0.norm2.bias",
+      "decoder.revealer_decoder.layers.0.norm3.weight",
+      "decoder.revealer_decoder.layers.0.norm3.bias",
+      "decoder.revealer_decoder.layers.1.self_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.1.self_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.1.self_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.1.self_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.1.multihead_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.1.multihead_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.1.multihead_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.1.multihead_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.1.linear1.weight",
+      "decoder.revealer_decoder.layers.1.linear1.bias",
+      "decoder.revealer_decoder.layers.1.linear2.weight",
+      "decoder.revealer_decoder.layers.1.linear2.bias",
+      "decoder.revealer_decoder.layers.1.norm1.weight",
+      "decoder.revealer_decoder.layers.1.norm1.bias",
+      "decoder.revealer_decoder.layers.1.norm2.weight",
+      "decoder.revealer_decoder.layers.1.norm2.bias",
+      "decoder.revealer_decoder.layers.1.norm3.weight",
+      "decoder.revealer_decoder.layers.1.norm3.bias",
+      "decoder.revealer_decoder.layers.2.self_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.2.self_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.2.self_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.2.self_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.2.multihead_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.2.multihead_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.2.multihead_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.2.multihead_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.2.linear1.weight",
+      "decoder.revealer_decoder.layers.2.linear1.bias",
+      "decoder.revealer_decoder.layers.2.linear2.weight",
+      "decoder.revealer_decoder.layers.2.linear2.bias",
+      "decoder.revealer_decoder.layers.2.norm1.weight",
+      "decoder.revealer_decoder.layers.2.norm1.bias",
+      "decoder.revealer_decoder.layers.2.norm2.weight",
+      "decoder.revealer_decoder.layers.2.norm2.bias",
+      "decoder.revealer_decoder.layers.2.norm3.weight",
+      "decoder.revealer_decoder.layers.2.norm3.bias",
+      "decoder.revealer_decoder.layers.3.self_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.3.self_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.3.self_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.3.self_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.3.multihead_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.3.multihead_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.3.multihead_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.3.multihead_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.3.linear1.weight",
+      "decoder.revealer_decoder.layers.3.linear1.bias",
+      "decoder.revealer_decoder.layers.3.linear2.weight",
+      "decoder.revealer_decoder.layers.3.linear2.bias",
+      "decoder.revealer_decoder.layers.3.norm1.weight",
+      "decoder.revealer_decoder.layers.3.norm1.bias",
+      "decoder.revealer_decoder.layers.3.norm2.weight",
+      "decoder.revealer_decoder.layers.3.norm2.bias",
+      "decoder.revealer_decoder.layers.3.norm3.weight",
+      "decoder.revealer_decoder.layers.3.norm3.bias",
+      "decoder.actor_decoder.layers.0.self_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.0.self_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.0.self_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.0.self_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.0.multihead_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.0.multihead_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.0.multihead_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.0.multihead_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.0.linear1.weight",
+      "decoder.actor_decoder.layers.0.linear1.bias",
+      "decoder.actor_decoder.layers.0.linear2.weight",
+      "decoder.actor_decoder.layers.0.linear2.bias",
+      "decoder.actor_decoder.layers.0.norm1.weight",
+      "decoder.actor_decoder.layers.0.norm1.bias",
+      "decoder.actor_decoder.layers.0.norm2.weight",
+      "decoder.actor_decoder.layers.0.norm2.bias",
+      "decoder.actor_decoder.layers.0.norm3.weight",
+      "decoder.actor_decoder.layers.0.norm3.bias",
+      "decoder.actor_decoder.layers.1.self_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.1.self_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.1.self_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.1.self_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.1.multihead_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.1.multihead_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.1.multihead_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.1.multihead_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.1.linear1.weight",
+      "decoder.actor_decoder.layers.1.linear1.bias",
+      "decoder.actor_decoder.layers.1.linear2.weight",
+      "decoder.actor_decoder.layers.1.linear2.bias",
+      "decoder.actor_decoder.layers.1.norm1.weight",
+      "decoder.actor_decoder.layers.1.norm1.bias",
+      "decoder.actor_decoder.layers.1.norm2.weight",
+      "decoder.actor_decoder.layers.1.norm2.bias",
+      "decoder.actor_decoder.layers.1.norm3.weight",
+      "decoder.actor_decoder.layers.1.norm3.bias",
+      "decoder.actor_decoder.layers.2.self_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.2.self_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.2.self_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.2.self_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.2.multihead_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.2.multihead_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.2.multihead_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.2.multihead_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.2.linear1.weight",
+      "decoder.actor_decoder.layers.2.linear1.bias",
+      "decoder.actor_decoder.layers.2.linear2.weight",
+      "decoder.actor_decoder.layers.2.linear2.bias",
+      "decoder.actor_decoder.layers.2.norm1.weight",
+      "decoder.actor_decoder.layers.2.norm1.bias",
+      "decoder.actor_decoder.layers.2.norm2.weight",
+      "decoder.actor_decoder.layers.2.norm2.bias",
+      "decoder.actor_decoder.layers.2.norm3.weight",
+      "decoder.actor_decoder.layers.2.norm3.bias",
+      "decoder.actor_decoder.layers.3.self_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.3.self_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.3.self_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.3.self_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.3.multihead_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.3.multihead_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.3.multihead_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.3.multihead_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.3.linear1.weight",
+      "decoder.actor_decoder.layers.3.linear1.bias",
+      "decoder.actor_decoder.layers.3.linear2.weight",
+      "decoder.actor_decoder.layers.3.linear2.bias",
+      "decoder.actor_decoder.layers.3.norm1.weight",
+      "decoder.actor_decoder.layers.3.norm1.bias",
+      "decoder.actor_decoder.layers.3.norm2.weight",
+      "decoder.actor_decoder.layers.3.norm2.bias",
+      "decoder.actor_decoder.layers.3.norm3.weight",
+      "decoder.actor_decoder.layers.3.norm3.bias",
+      "decoder.revealer_mean.weight",
+      "decoder.revealer_mean.bias",
+      "decoder.revealer_log_std.weight",
+      "decoder.revealer_log_std.bias",
+      "decoder.actor_mean.weight",
+      "decoder.actor_mean.bias",
+      "decoder.actor_log_std.weight",
+      "decoder.actor_log_std.bias",
+      "decoder.proposal_score.0.weight",
+      "decoder.proposal_score.0.bias",
+      "decoder.proposal_score.1.weight",
+      "decoder.proposal_score.1.bias"
+    ],
+    "unexpected_keys": []
+  },
+  "trainable_parameter_names": [
+    "fusion.camera_embedding.weight",
+    "fusion.cross_view_transformer.layers.0.self_attn.in_proj_weight",
+    "fusion.cross_view_transformer.layers.0.self_attn.in_proj_bias",
+    "fusion.cross_view_transformer.layers.0.self_attn.out_proj.weight",
+    "fusion.cross_view_transformer.layers.0.self_attn.out_proj.bias",
+    "fusion.cross_view_transformer.layers.0.linear1.weight",
+    "fusion.cross_view_transformer.layers.0.linear1.bias",
+    "fusion.cross_view_transformer.layers.0.linear2.weight",
+    "fusion.cross_view_transformer.layers.0.linear2.bias",
+    "fusion.cross_view_transformer.layers.0.norm1.weight",
+    "fusion.cross_view_transformer.layers.0.norm1.bias",
+    "fusion.cross_view_transformer.layers.0.norm2.weight",
+    "fusion.cross_view_transformer.layers.0.norm2.bias",
+    "fusion.cross_view_transformer.layers.1.self_attn.in_proj_weight",
+    "fusion.cross_view_transformer.layers.1.self_attn.in_proj_bias",
+    "fusion.cross_view_transformer.layers.1.self_attn.out_proj.weight",
+    "fusion.cross_view_transformer.layers.1.self_attn.out_proj.bias",
+    "fusion.cross_view_transformer.layers.1.linear1.weight",
+    "fusion.cross_view_transformer.layers.1.linear1.bias",
+    "fusion.cross_view_transformer.layers.1.linear2.weight",
+    "fusion.cross_view_transformer.layers.1.linear2.bias",
+    "fusion.cross_view_transformer.layers.1.norm1.weight",
+    "fusion.cross_view_transformer.layers.1.norm1.bias",
+    "fusion.cross_view_transformer.layers.1.norm2.weight",
+    "fusion.cross_view_transformer.layers.1.norm2.bias",
+    "fusion.cross_view_transformer.layers.2.self_attn.in_proj_weight",
+    "fusion.cross_view_transformer.layers.2.self_attn.in_proj_bias",
+    "fusion.cross_view_transformer.layers.2.self_attn.out_proj.weight",
+    "fusion.cross_view_transformer.layers.2.self_attn.out_proj.bias",
+    "fusion.cross_view_transformer.layers.2.linear1.weight",
+    "fusion.cross_view_transformer.layers.2.linear1.bias",
+    "fusion.cross_view_transformer.layers.2.linear2.weight",
+    "fusion.cross_view_transformer.layers.2.linear2.bias",
+    "fusion.cross_view_transformer.layers.2.norm1.weight",
+    "fusion.cross_view_transformer.layers.2.norm1.bias",
+    "fusion.cross_view_transformer.layers.2.norm2.weight",
+    "fusion.cross_view_transformer.layers.2.norm2.bias",
+    "fusion.cross_view_transformer.layers.3.self_attn.in_proj_weight",
+    "fusion.cross_view_transformer.layers.3.self_attn.in_proj_bias",
+    "fusion.cross_view_transformer.layers.3.self_attn.out_proj.weight",
+    "fusion.cross_view_transformer.layers.3.self_attn.out_proj.bias",
+    "fusion.cross_view_transformer.layers.3.linear1.weight",
+    "fusion.cross_view_transformer.layers.3.linear1.bias",
+    "fusion.cross_view_transformer.layers.3.linear2.weight",
+    "fusion.cross_view_transformer.layers.3.linear2.bias",
+    "fusion.cross_view_transformer.layers.3.norm1.weight",
+    "fusion.cross_view_transformer.layers.3.norm1.bias",
+    "fusion.cross_view_transformer.layers.3.norm2.weight",
+    "fusion.cross_view_transformer.layers.3.norm2.bias",
+    "fusion.geometry_fusion.attn.in_proj_weight",
+    "fusion.geometry_fusion.attn.in_proj_bias",
+    "fusion.geometry_fusion.attn.out_proj.weight",
+    "fusion.geometry_fusion.attn.out_proj.bias",
+    "fusion.geometry_fusion.gate.0.weight",
+    "fusion.geometry_fusion.gate.0.bias",
+    "fusion.geometry_fusion.gate.1.weight",
+    "fusion.geometry_fusion.gate.1.bias",
+    "fusion.geometry_fusion.gate.3.weight",
+    "fusion.geometry_fusion.gate.3.bias",
+    "fusion.geometry_fusion.out.0.weight",
+    "fusion.geometry_fusion.out.0.bias",
+    "fusion.geometry_fusion.out.1.weight",
+    "fusion.geometry_fusion.out.1.bias",
+    "fusion.proprio_adapter.0.weight",
+    "fusion.proprio_adapter.0.bias",
+    "fusion.proprio_adapter.1.weight",
+    "fusion.proprio_adapter.1.bias",
+    "memory.gru.weight_ih_l0",
+    "memory.gru.weight_hh_l0",
+    "memory.gru.bias_ih_l0",
+    "memory.gru.bias_hh_l0",
+    "memory.gru.weight_ih_l1",
+    "memory.gru.weight_hh_l1",
+    "memory.gru.bias_ih_l1",
+    "memory.gru.bias_hh_l1",
+    "memory.token_proj.0.weight",
+    "memory.token_proj.0.bias",
+    "memory.token_proj.1.weight",
+    "memory.token_proj.1.bias",
+    "memory.action_proj.0.weight",
+    "memory.action_proj.0.bias",
+    "memory.action_proj.1.weight",
+    "memory.action_proj.1.bias",
+    "memory.uncertainty_head.0.weight",
+    "memory.uncertainty_head.0.bias",
+    "memory.uncertainty_head.1.weight",
+    "memory.uncertainty_head.1.bias",
+    "decoder.actor_role_bias",
+    "decoder.revealer_decoder.layers.0.self_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.0.self_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.0.self_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.0.self_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.0.multihead_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.0.multihead_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.0.multihead_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.0.multihead_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.0.linear1.weight",
+    "decoder.revealer_decoder.layers.0.linear1.bias",
+    "decoder.revealer_decoder.layers.0.linear2.weight",
+    "decoder.revealer_decoder.layers.0.linear2.bias",
+    "decoder.revealer_decoder.layers.0.norm1.weight",
+    "decoder.revealer_decoder.layers.0.norm1.bias",
+    "decoder.revealer_decoder.layers.0.norm2.weight",
+    "decoder.revealer_decoder.layers.0.norm2.bias",
+    "decoder.revealer_decoder.layers.0.norm3.weight",
+    "decoder.revealer_decoder.layers.0.norm3.bias",
+    "decoder.revealer_decoder.layers.1.self_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.1.self_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.1.self_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.1.self_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.1.multihead_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.1.multihead_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.1.multihead_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.1.multihead_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.1.linear1.weight",
+    "decoder.revealer_decoder.layers.1.linear1.bias",
+    "decoder.revealer_decoder.layers.1.linear2.weight",
+    "decoder.revealer_decoder.layers.1.linear2.bias",
+    "decoder.revealer_decoder.layers.1.norm1.weight",
+    "decoder.revealer_decoder.layers.1.norm1.bias",
+    "decoder.revealer_decoder.layers.1.norm2.weight",
+    "decoder.revealer_decoder.layers.1.norm2.bias",
+    "decoder.revealer_decoder.layers.1.norm3.weight",
+    "decoder.revealer_decoder.layers.1.norm3.bias",
+    "decoder.revealer_decoder.layers.2.self_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.2.self_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.2.self_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.2.self_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.2.multihead_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.2.multihead_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.2.multihead_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.2.multihead_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.2.linear1.weight",
+    "decoder.revealer_decoder.layers.2.linear1.bias",
+    "decoder.revealer_decoder.layers.2.linear2.weight",
+    "decoder.revealer_decoder.layers.2.linear2.bias",
+    "decoder.revealer_decoder.layers.2.norm1.weight",
+    "decoder.revealer_decoder.layers.2.norm1.bias",
+    "decoder.revealer_decoder.layers.2.norm2.weight",
+    "decoder.revealer_decoder.layers.2.norm2.bias",
+    "decoder.revealer_decoder.layers.2.norm3.weight",
+    "decoder.revealer_decoder.layers.2.norm3.bias",
+    "decoder.revealer_decoder.layers.3.self_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.3.self_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.3.self_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.3.self_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.3.multihead_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.3.multihead_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.3.multihead_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.3.multihead_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.3.linear1.weight",
+    "decoder.revealer_decoder.layers.3.linear1.bias",
+    "decoder.revealer_decoder.layers.3.linear2.weight",
+    "decoder.revealer_decoder.layers.3.linear2.bias",
+    "decoder.revealer_decoder.layers.3.norm1.weight",
+    "decoder.revealer_decoder.layers.3.norm1.bias",
+    "decoder.revealer_decoder.layers.3.norm2.weight",
+    "decoder.revealer_decoder.layers.3.norm2.bias",
+    "decoder.revealer_decoder.layers.3.norm3.weight",
+    "decoder.revealer_decoder.layers.3.norm3.bias",
+    "decoder.actor_decoder.layers.0.self_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.0.self_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.0.self_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.0.self_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.0.multihead_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.0.multihead_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.0.multihead_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.0.multihead_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.0.linear1.weight",
+    "decoder.actor_decoder.layers.0.linear1.bias",
+    "decoder.actor_decoder.layers.0.linear2.weight",
+    "decoder.actor_decoder.layers.0.linear2.bias",
+    "decoder.actor_decoder.layers.0.norm1.weight",
+    "decoder.actor_decoder.layers.0.norm1.bias",
+    "decoder.actor_decoder.layers.0.norm2.weight",
+    "decoder.actor_decoder.layers.0.norm2.bias",
+    "decoder.actor_decoder.layers.0.norm3.weight",
+    "decoder.actor_decoder.layers.0.norm3.bias",
+    "decoder.actor_decoder.layers.1.self_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.1.self_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.1.self_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.1.self_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.1.multihead_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.1.multihead_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.1.multihead_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.1.multihead_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.1.linear1.weight",
+    "decoder.actor_decoder.layers.1.linear1.bias",
+    "decoder.actor_decoder.layers.1.linear2.weight",
+    "decoder.actor_decoder.layers.1.linear2.bias",
+    "decoder.actor_decoder.layers.1.norm1.weight",
+    "decoder.actor_decoder.layers.1.norm1.bias",
+    "decoder.actor_decoder.layers.1.norm2.weight",
+    "decoder.actor_decoder.layers.1.norm2.bias",
+    "decoder.actor_decoder.layers.1.norm3.weight",
+    "decoder.actor_decoder.layers.1.norm3.bias",
+    "decoder.actor_decoder.layers.2.self_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.2.self_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.2.self_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.2.self_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.2.multihead_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.2.multihead_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.2.multihead_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.2.multihead_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.2.linear1.weight",
+    "decoder.actor_decoder.layers.2.linear1.bias",
+    "decoder.actor_decoder.layers.2.linear2.weight",
+    "decoder.actor_decoder.layers.2.linear2.bias",
+    "decoder.actor_decoder.layers.2.norm1.weight",
+    "decoder.actor_decoder.layers.2.norm1.bias",
+    "decoder.actor_decoder.layers.2.norm2.weight",
+    "decoder.actor_decoder.layers.2.norm2.bias",
+    "decoder.actor_decoder.layers.2.norm3.weight",
+    "decoder.actor_decoder.layers.2.norm3.bias",
+    "decoder.actor_decoder.layers.3.self_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.3.self_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.3.self_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.3.self_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.3.multihead_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.3.multihead_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.3.multihead_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.3.multihead_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.3.linear1.weight",
+    "decoder.actor_decoder.layers.3.linear1.bias",
+    "decoder.actor_decoder.layers.3.linear2.weight",
+    "decoder.actor_decoder.layers.3.linear2.bias",
+    "decoder.actor_decoder.layers.3.norm1.weight",
+    "decoder.actor_decoder.layers.3.norm1.bias",
+    "decoder.actor_decoder.layers.3.norm2.weight",
+    "decoder.actor_decoder.layers.3.norm2.bias",
+    "decoder.actor_decoder.layers.3.norm3.weight",
+    "decoder.actor_decoder.layers.3.norm3.bias",
+    "decoder.query_embed.weight",
+    "decoder.revealer_mean.weight",
+    "decoder.revealer_mean.bias",
+    "decoder.revealer_log_std.weight",
+    "decoder.revealer_log_std.bias",
+    "decoder.actor_mean.weight",
+    "decoder.actor_mean.bias",
+    "decoder.actor_log_std.weight",
+    "decoder.actor_log_std.bias",
+    "decoder.coordination.0.weight",
+    "decoder.coordination.0.bias",
+    "decoder.coordination.1.weight",
+    "decoder.coordination.1.bias",
+    "decoder.coordination.3.weight",
+    "decoder.coordination.3.bias",
+    "decoder.proposal_score.0.weight",
+    "decoder.proposal_score.0.bias",
+    "decoder.proposal_score.1.weight",
+    "decoder.proposal_score.1.bias"
+  ],
+  "best_val_total": 0.37670365827424185,
+  "history": [
+    {
+      "epoch": 0,
+      "train": {
+        "action": 0.6513718327409342,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.2277212582136456,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.7679335788676613,
+        "role_swap_consistency": 0.0,
+        "total": 0.9276820499646036,
+        "transition": 0.0,
+        "world_model": 0.0
+      },
+      "val": {
+        "action": 0.16291735001972743,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.1196258578981673,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.6713550175939288,
+        "role_swap_consistency": 0.0,
+        "total": 0.41142383217811584,
+        "transition": 0.0,
+        "world_model": 0.0
+      }
+    },
+    {
+      "epoch": 1,
+      "train": {
+        "action": 0.21071406963624453,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.155882295809294,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.6945144220402366,
+        "role_swap_consistency": 0.0,
+        "total": 0.467438153530422,
+        "transition": 0.0,
+        "world_model": 0.0
+      },
+      "val": {
+        "action": 0.23247837594577245,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.1327585322516305,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.6880117058753967,
+        "role_swap_consistency": 0.0,
+        "total": 0.48495357377188547,
+        "transition": 0.0,
+        "world_model": 0.0
+      }
+    },
+    {
+      "epoch": 2,
+      "train": {
+        "action": 0.1608393647168812,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.2125610175885653,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.7512044781132748,
+        "role_swap_consistency": 0.0,
+        "total": 0.4328680618813163,
+        "transition": 0.0,
+        "world_model": 0.0
+      },
+      "val": {
+        "action": 0.09432156596864973,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.305886251585824,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.798845146383558,
+        "role_swap_consistency": 0.0,
+        "total": 0.3860659216131483,
+        "transition": 0.0,
+        "world_model": 0.0
+      }
+    },
+    {
+      "epoch": 3,
+      "train": {
+        "action": 0.1447360997921542,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.175407836311742,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.7467456679595145,
+        "role_swap_consistency": 0.0,
+        "total": 0.4106567674561551,
+        "transition": 0.0,
+        "world_model": 0.0
+      },
+      "val": {
+        "action": 0.13610392383166722,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.1333176749093192,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.6300536692142487,
+        "role_swap_consistency": 0.0,
+        "total": 0.38170802167483736,
+        "transition": 0.0,
+        "world_model": 0.0
+      }
+    },
+    {
+      "epoch": 4,
+      "train": {
+        "action": 0.13893984062106987,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.2317605771516498,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.7522393684638174,
+        "role_swap_consistency": 0.0,
+        "total": 0.4139726554092608,
+        "transition": 0.0,
+        "world_model": 0.0
+      },
+      "val": {
+        "action": 0.08872867269175393,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.3585667099271501,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.8373270290238517,
+        "role_swap_consistency": 0.0,
+        "total": 0.39299292649541584,
+        "transition": 0.0,
+        "world_model": 0.0
+      }
+    },
+    {
+      "epoch": 5,
+      "train": {
+        "action": 0.14546634727402738,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.1457386205070896,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.6874889543181971,
+        "role_swap_consistency": 0.0,
+        "total": 0.3998258223659114,
+        "transition": 0.0,
+        "world_model": 0.0
+      },
+      "val": {
+        "action": 0.13263668226344244,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.099511010306222,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.6595027276447841,
+        "role_swap_consistency": 0.0,
+        "total": 0.37670365827424185,
+        "transition": 0.0,
+        "world_model": 0.0
+      }
+    }
+  ],
+  "train_spec": {
+    "track_id": "bag_track",
+    "suite": "maniskill3",
+    "benchmark_task": "PutEggplantInBasketRetrievalProxy-v1",
+    "model_variant": "trunk_only_ft",
+    "seed": 17,
+    "train_demos": 32,
+    "val_demos": 8,
+    "init_checkpoint_group": "/workspace/workspace/VLAarchtests2/VLAarchtests/artifacts/outputs/r3d_handoff_phase/proxy_interaction_r3d_stage3_clip_rgbd_handoff_compact_phase_seed17/checkpoint_best.pt",
+    "optimizer": "adamw",
+    "learning_rate": 0.0001,
+    "lr_schedule": "constant",
+    "batch_size": 4,
+    "augmentations": "none",
+    "early_stopping_metric": "val_total",
+    "max_gradient_steps": 114,
+    "unfreeze_scope": "fusion_memory_decoder",
+    "dataset_split_id": "bag_bridge_smoke_v1_seed17",
+    "same_data_policy": true,
+    "same_init_policy": true
+  }
+}

outputs/maniskill_bag_bridge_smoke_v1/trunk_only_ft_seed23/checkpoint_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9245c09caccaea670d910bf86d60a35d7db1d345cdc71d55a4e854d903ce0b3
+size 817966700

outputs/maniskill_bag_bridge_smoke_v1/trunk_only_ft_seed23/summary.json

@@ -0,0 +1,1030 @@
+{
+  "task": "bag",
+  "variant": "trunk_only_ft",
+  "checkpoint_path": "/workspace/workspace/outputs/maniskill_bag_bridge_smoke_v1/trunk_only_ft_seed23/checkpoint_best.pt",
+  "init_info": {
+    "path": "/workspace/workspace/VLAarchtests2/VLAarchtests/artifacts/outputs/r3d_handoff_phase/proxy_interaction_r3d_stage3_clip_rgbd_handoff_compact_phase_seed17/checkpoint_best.pt",
+    "loaded_keys": 489,
+    "skipped_shape_mismatch_keys": [
+      "memory.scene_memory.position_embedding",
+      "memory.scene_memory.bank_queries",
+      "memory.scene_memory.sequence_encoder.layers.0.self_attn.in_proj_weight",
+      "memory.scene_memory.sequence_encoder.layers.0.self_attn.in_proj_bias",
+      "memory.scene_memory.sequence_encoder.layers.0.self_attn.out_proj.weight",
+      "memory.scene_memory.sequence_encoder.layers.0.self_attn.out_proj.bias",
+      "memory.scene_memory.sequence_encoder.layers.0.linear1.weight",
+      "memory.scene_memory.sequence_encoder.layers.0.linear1.bias",
+      "memory.scene_memory.sequence_encoder.layers.0.linear2.weight",
+      "memory.scene_memory.sequence_encoder.layers.0.linear2.bias",
+      "memory.scene_memory.sequence_encoder.layers.0.norm1.weight",
+      "memory.scene_memory.sequence_encoder.layers.0.norm1.bias",
+      "memory.scene_memory.sequence_encoder.layers.0.norm2.weight",
+      "memory.scene_memory.sequence_encoder.layers.0.norm2.bias",
+      "memory.scene_memory.bank_attention.in_proj_weight",
+      "memory.scene_memory.bank_attention.in_proj_bias",
+      "memory.scene_memory.bank_attention.out_proj.weight",
+      "memory.scene_memory.bank_attention.out_proj.bias",
+      "memory.scene_memory.action_proj.0.weight",
+      "memory.scene_memory.action_proj.0.bias",
+      "memory.scene_memory.action_proj.1.weight",
+      "memory.scene_memory.action_proj.1.bias",
+      "memory.scene_memory.write_gate.0.weight",
+      "memory.scene_memory.write_gate.0.bias",
+      "memory.scene_memory.write_gate.1.weight",
+      "memory.scene_memory.write_gate.1.bias",
+      "memory.scene_memory.write_gate.3.weight",
+      "memory.scene_memory.write_gate.3.bias",
+      "memory.scene_memory.token_proj.0.weight",
+      "memory.scene_memory.token_proj.0.bias",
+      "memory.scene_memory.token_proj.1.weight",
+      "memory.scene_memory.token_proj.1.bias",
+      "memory.belief_memory.position_embedding",
+      "memory.belief_memory.bank_queries",
+      "memory.belief_memory.sequence_encoder.layers.0.self_attn.in_proj_weight",
+      "memory.belief_memory.sequence_encoder.layers.0.self_attn.in_proj_bias",
+      "memory.belief_memory.sequence_encoder.layers.0.self_attn.out_proj.weight",
+      "memory.belief_memory.sequence_encoder.layers.0.self_attn.out_proj.bias",
+      "memory.belief_memory.sequence_encoder.layers.0.linear1.weight",
+      "memory.belief_memory.sequence_encoder.layers.0.linear1.bias",
+      "memory.belief_memory.sequence_encoder.layers.0.linear2.weight",
+      "memory.belief_memory.sequence_encoder.layers.0.linear2.bias",
+      "memory.belief_memory.sequence_encoder.layers.0.norm1.weight",
+      "memory.belief_memory.sequence_encoder.layers.0.norm1.bias",
+      "memory.belief_memory.sequence_encoder.layers.0.norm2.weight",
+      "memory.belief_memory.sequence_encoder.layers.0.norm2.bias",
+      "memory.belief_memory.bank_attention.in_proj_weight",
+      "memory.belief_memory.bank_attention.in_proj_bias",
+      "memory.belief_memory.bank_attention.out_proj.weight",
+      "memory.belief_memory.bank_attention.out_proj.bias",
+      "memory.belief_memory.action_proj.0.weight",
+      "memory.belief_memory.action_proj.0.bias",
+      "memory.belief_memory.action_proj.1.weight",
+      "memory.belief_memory.action_proj.1.bias",
+      "memory.belief_memory.write_gate.0.weight",
+      "memory.belief_memory.write_gate.0.bias",
+      "memory.belief_memory.write_gate.1.weight",
+      "memory.belief_memory.write_gate.1.bias",
+      "memory.belief_memory.write_gate.3.weight",
+      "memory.belief_memory.write_gate.3.bias",
+      "memory.belief_memory.token_proj.0.weight",
+      "memory.belief_memory.token_proj.0.bias",
+      "memory.belief_memory.token_proj.1.weight",
+      "memory.belief_memory.token_proj.1.bias",
+      "decoder.arm_decoder.layers.0.self_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.0.self_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.0.self_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.0.self_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.0.multihead_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.0.multihead_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.0.multihead_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.0.multihead_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.0.linear1.weight",
+      "decoder.arm_decoder.layers.0.linear1.bias",
+      "decoder.arm_decoder.layers.0.linear2.weight",
+      "decoder.arm_decoder.layers.0.linear2.bias",
+      "decoder.arm_decoder.layers.0.norm1.weight",
+      "decoder.arm_decoder.layers.0.norm1.bias",
+      "decoder.arm_decoder.layers.0.norm2.weight",
+      "decoder.arm_decoder.layers.0.norm2.bias",
+      "decoder.arm_decoder.layers.0.norm3.weight",
+      "decoder.arm_decoder.layers.0.norm3.bias",
+      "decoder.arm_decoder.layers.1.self_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.1.self_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.1.self_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.1.self_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.1.multihead_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.1.multihead_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.1.multihead_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.1.multihead_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.1.linear1.weight",
+      "decoder.arm_decoder.layers.1.linear1.bias",
+      "decoder.arm_decoder.layers.1.linear2.weight",
+      "decoder.arm_decoder.layers.1.linear2.bias",
+      "decoder.arm_decoder.layers.1.norm1.weight",
+      "decoder.arm_decoder.layers.1.norm1.bias",
+      "decoder.arm_decoder.layers.1.norm2.weight",
+      "decoder.arm_decoder.layers.1.norm2.bias",
+      "decoder.arm_decoder.layers.1.norm3.weight",
+      "decoder.arm_decoder.layers.1.norm3.bias",
+      "decoder.arm_decoder.layers.2.self_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.2.self_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.2.self_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.2.self_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.2.multihead_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.2.multihead_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.2.multihead_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.2.multihead_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.2.linear1.weight",
+      "decoder.arm_decoder.layers.2.linear1.bias",
+      "decoder.arm_decoder.layers.2.linear2.weight",
+      "decoder.arm_decoder.layers.2.linear2.bias",
+      "decoder.arm_decoder.layers.2.norm1.weight",
+      "decoder.arm_decoder.layers.2.norm1.bias",
+      "decoder.arm_decoder.layers.2.norm2.weight",
+      "decoder.arm_decoder.layers.2.norm2.bias",
+      "decoder.arm_decoder.layers.2.norm3.weight",
+      "decoder.arm_decoder.layers.2.norm3.bias",
+      "decoder.arm_decoder.layers.3.self_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.3.self_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.3.self_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.3.self_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.3.multihead_attn.in_proj_weight",
+      "decoder.arm_decoder.layers.3.multihead_attn.in_proj_bias",
+      "decoder.arm_decoder.layers.3.multihead_attn.out_proj.weight",
+      "decoder.arm_decoder.layers.3.multihead_attn.out_proj.bias",
+      "decoder.arm_decoder.layers.3.linear1.weight",
+      "decoder.arm_decoder.layers.3.linear1.bias",
+      "decoder.arm_decoder.layers.3.linear2.weight",
+      "decoder.arm_decoder.layers.3.linear2.bias",
+      "decoder.arm_decoder.layers.3.norm1.weight",
+      "decoder.arm_decoder.layers.3.norm1.bias",
+      "decoder.arm_decoder.layers.3.norm2.weight",
+      "decoder.arm_decoder.layers.3.norm2.bias",
+      "decoder.arm_decoder.layers.3.norm3.weight",
+      "decoder.arm_decoder.layers.3.norm3.bias",
+      "decoder.arm_identity.weight",
+      "decoder.task_embedding.weight",
+      "decoder.phase_adapter.weight",
+      "decoder.phase_adapter.bias",
+      "decoder.role_adapter.weight",
+      "decoder.role_adapter.bias",
+      "decoder.context_proj.0.weight",
+      "decoder.context_proj.0.bias",
+      "decoder.context_proj.1.weight",
+      "decoder.context_proj.1.bias",
+      "decoder.arm_head.0.weight",
+      "decoder.arm_head.0.bias",
+      "decoder.arm_head.1.weight",
+      "decoder.arm_head.1.bias",
+      "decoder.arm_mean.weight",
+      "decoder.arm_mean.bias",
+      "decoder.arm_log_std.weight",
+      "decoder.arm_log_std.bias",
+      "decoder.proposal_mode_head.0.weight",
+      "decoder.proposal_mode_head.0.bias",
+      "decoder.proposal_mode_head.1.weight",
+      "decoder.proposal_mode_head.1.bias",
+      "decoder.proposal_mode_head.3.weight",
+      "decoder.proposal_mode_head.3.bias",
+      "decoder.proposal_mode_embeddings.weight",
+      "decoder.proposal_slot_embeddings.weight",
+      "decoder.mode_residual_heads.0.0.weight",
+      "decoder.mode_residual_heads.0.0.bias",
+      "decoder.mode_residual_heads.0.1.weight",
+      "decoder.mode_residual_heads.0.1.bias",
+      "decoder.mode_residual_heads.0.3.weight",
+      "decoder.mode_residual_heads.0.3.bias",
+      "decoder.mode_residual_heads.1.0.weight",
+      "decoder.mode_residual_heads.1.0.bias",
+      "decoder.mode_residual_heads.1.1.weight",
+      "decoder.mode_residual_heads.1.1.bias",
+      "decoder.mode_residual_heads.1.3.weight",
+      "decoder.mode_residual_heads.1.3.bias",
+      "decoder.mode_residual_heads.2.0.weight",
+      "decoder.mode_residual_heads.2.0.bias",
+      "decoder.mode_residual_heads.2.1.weight",
+      "decoder.mode_residual_heads.2.1.bias",
+      "decoder.mode_residual_heads.2.3.weight",
+      "decoder.mode_residual_heads.2.3.bias",
+      "decoder.mode_residual_heads.3.0.weight",
+      "decoder.mode_residual_heads.3.0.bias",
+      "decoder.mode_residual_heads.3.1.weight",
+      "decoder.mode_residual_heads.3.1.bias",
+      "decoder.mode_residual_heads.3.3.weight",
+      "decoder.mode_residual_heads.3.3.bias",
+      "decoder.mode_residual_heads.4.0.weight",
+      "decoder.mode_residual_heads.4.0.bias",
+      "decoder.mode_residual_heads.4.1.weight",
+      "decoder.mode_residual_heads.4.1.bias",
+      "decoder.mode_residual_heads.4.3.weight",
+      "decoder.mode_residual_heads.4.3.bias",
+      "decoder.mode_residual_heads.5.0.weight",
+      "decoder.mode_residual_heads.5.0.bias",
+      "decoder.mode_residual_heads.5.1.weight",
+      "decoder.mode_residual_heads.5.1.bias",
+      "decoder.mode_residual_heads.5.3.weight",
+      "decoder.mode_residual_heads.5.3.bias",
+      "decoder.mode_residual_heads.6.0.weight",
+      "decoder.mode_residual_heads.6.0.bias",
+      "decoder.mode_residual_heads.6.1.weight",
+      "decoder.mode_residual_heads.6.1.bias",
+      "decoder.mode_residual_heads.6.3.weight",
+      "decoder.mode_residual_heads.6.3.bias",
+      "decoder.slot_delta.0.weight",
+      "decoder.slot_delta.0.bias",
+      "decoder.slot_delta.1.weight",
+      "decoder.slot_delta.1.bias",
+      "decoder.slot_delta.3.weight",
+      "decoder.slot_delta.3.bias",
+      "decoder.proposal_score.0.weight",
+      "decoder.proposal_score.0.bias",
+      "decoder.proposal_score.1.weight",
+      "decoder.proposal_score.1.bias",
+      "decoder.proposal_score.3.weight",
+      "decoder.proposal_score.3.bias",
+      "elastic_state_head.interaction_queries",
+      "elastic_state_head.interaction_attention.in_proj_weight",
+      "elastic_state_head.interaction_attention.in_proj_bias",
+      "elastic_state_head.interaction_attention.out_proj.weight",
+      "elastic_state_head.interaction_attention.out_proj.bias",
+      "elastic_state_head.interaction_mlp.0.weight",
+      "elastic_state_head.interaction_mlp.0.bias",
+      "elastic_state_head.interaction_mlp.1.weight",
+      "elastic_state_head.interaction_mlp.1.bias",
+      "elastic_state_head.interaction_mlp.3.weight",
+      "elastic_state_head.interaction_mlp.3.bias",
+      "elastic_state_head.decoder.field_queries",
+      "elastic_state_head.decoder.field_attention.in_proj_weight",
+      "elastic_state_head.decoder.field_attention.in_proj_bias",
+      "elastic_state_head.decoder.field_attention.out_proj.weight",
+      "elastic_state_head.decoder.field_attention.out_proj.bias",
+      "elastic_state_head.decoder.field_mlp.0.weight",
+      "elastic_state_head.decoder.field_mlp.0.bias",
+      "elastic_state_head.decoder.field_mlp.1.weight",
+      "elastic_state_head.decoder.field_mlp.1.bias",
+      "elastic_state_head.decoder.field_mlp.3.weight",
+      "elastic_state_head.decoder.field_mlp.3.bias",
+      "elastic_state_head.decoder.summary_proj.0.weight",
+      "elastic_state_head.decoder.summary_proj.0.bias",
+      "elastic_state_head.decoder.summary_proj.1.weight",
+      "elastic_state_head.decoder.summary_proj.1.bias",
+      "elastic_state_head.decoder.phase_head.0.weight",
+      "elastic_state_head.decoder.phase_head.0.bias",
+      "elastic_state_head.decoder.phase_head.1.weight",
+      "elastic_state_head.decoder.phase_head.1.bias",
+      "elastic_state_head.decoder.phase_head.3.weight",
+      "elastic_state_head.decoder.phase_head.3.bias",
+      "elastic_state_head.decoder.arm_role_head.0.weight",
+      "elastic_state_head.decoder.arm_role_head.0.bias",
+      "elastic_state_head.decoder.arm_role_head.1.weight",
+      "elastic_state_head.decoder.arm_role_head.1.bias",
+      "elastic_state_head.decoder.arm_role_head.3.weight",
+      "elastic_state_head.decoder.arm_role_head.3.bias",
+      "elastic_state_head.decoder.arm_identity.weight",
+      "elastic_state_head.decoder.support_mode.0.weight",
+      "elastic_state_head.decoder.support_mode.0.bias",
+      "elastic_state_head.decoder.support_mode.1.weight",
+      "elastic_state_head.decoder.support_mode.1.bias",
+      "elastic_state_head.decoder.support_mode.3.weight",
+      "elastic_state_head.decoder.support_mode.3.bias",
+      "elastic_state_head.decoder.access_field.weight",
+      "elastic_state_head.decoder.access_field.bias",
+      "elastic_state_head.decoder.target_belief_field.weight",
+      "elastic_state_head.decoder.target_belief_field.bias",
+      "elastic_state_head.decoder.visibility_field.weight",
+      "elastic_state_head.decoder.visibility_field.bias",
+      "elastic_state_head.decoder.clearance_field.weight",
+      "elastic_state_head.decoder.clearance_field.bias",
+      "elastic_state_head.decoder.occluder_contact_field.weight",
+      "elastic_state_head.decoder.occluder_contact_field.bias",
+      "elastic_state_head.decoder.grasp_affordance_field.weight",
+      "elastic_state_head.decoder.grasp_affordance_field.bias",
+      "elastic_state_head.decoder.support_stability_field.weight",
+      "elastic_state_head.decoder.support_stability_field.bias",
+      "elastic_state_head.decoder.persistence_field.weight",
+      "elastic_state_head.decoder.persistence_field.bias",
+      "elastic_state_head.decoder.reocclusion_field.weight",
+      "elastic_state_head.decoder.reocclusion_field.bias",
+      "elastic_state_head.decoder.disturbance_field.weight",
+      "elastic_state_head.decoder.disturbance_field.bias",
+      "elastic_state_head.decoder.uncertainty_field.weight",
+      "elastic_state_head.decoder.uncertainty_field.bias",
+      "elastic_state_head.decoder.reocclusion_head.0.weight",
+      "elastic_state_head.decoder.reocclusion_head.0.bias",
+      "elastic_state_head.decoder.reocclusion_head.1.weight",
+      "elastic_state_head.decoder.reocclusion_head.1.bias",
+      "elastic_state_head.decoder.reocclusion_head.3.weight",
+      "elastic_state_head.decoder.reocclusion_head.3.bias",
+      "elastic_state_head.decoder.task_embedding.weight",
+      "elastic_state_head.decoder.task_field_affine.weight",
+      "elastic_state_head.decoder.task_field_affine.bias",
+      "elastic_state_head.decoder.task_summary_adapter.0.weight",
+      "elastic_state_head.decoder.task_summary_adapter.0.bias",
+      "elastic_state_head.decoder.task_summary_adapter.1.weight",
+      "elastic_state_head.decoder.task_summary_adapter.1.bias",
+      "elastic_state_head.decoder.task_phase_head.weight",
+      "elastic_state_head.decoder.task_phase_head.bias",
+      "elastic_state_head.decoder.task_support_head.weight",
+      "elastic_state_head.decoder.task_support_head.bias",
+      "elastic_state_head.decoder.task_reocclusion_head.weight",
+      "elastic_state_head.decoder.task_reocclusion_head.bias",
+      "elastic_state_head.decoder.task_metric_head.0.weight",
+      "elastic_state_head.decoder.task_metric_head.0.bias",
+      "elastic_state_head.decoder.task_metric_head.1.weight",
+      "elastic_state_head.decoder.task_metric_head.1.bias",
+      "elastic_state_head.decoder.task_metric_head.3.weight",
+      "elastic_state_head.decoder.task_metric_head.3.bias",
+      "world_model.state_encoder.0.weight",
+      "world_model.state_encoder.0.bias",
+      "world_model.state_encoder.1.weight",
+      "world_model.state_encoder.1.bias",
+      "world_model.scene_memory_proj.0.weight",
+      "world_model.scene_memory_proj.0.bias",
+      "world_model.scene_memory_proj.1.weight",
+      "world_model.scene_memory_proj.1.bias",
+      "world_model.belief_memory_proj.0.weight",
+      "world_model.belief_memory_proj.0.bias",
+      "world_model.belief_memory_proj.1.weight",
+      "world_model.belief_memory_proj.1.bias",
+      "world_model.action_encoder.0.weight",
+      "world_model.action_encoder.0.bias",
+      "world_model.action_encoder.1.weight",
+      "world_model.action_encoder.1.bias",
+      "world_model.transition.weight_ih",
+      "world_model.transition.weight_hh",
+      "world_model.transition.bias_ih",
+      "world_model.transition.bias_hh",
+      "world_model.scene_memory_update.weight",
+      "world_model.scene_memory_update.bias",
+      "world_model.belief_memory_update.weight",
+      "world_model.belief_memory_update.bias",
+      "world_model.compact_decoder.weight",
+      "world_model.compact_decoder.bias",
+      "world_model.target_belief_head.weight",
+      "world_model.target_belief_head.bias",
+      "world_model.visibility_head.weight",
+      "world_model.visibility_head.bias",
+      "world_model.clearance_head.weight",
+      "world_model.clearance_head.bias",
+      "world_model.occluder_contact_head.weight",
+      "world_model.occluder_contact_head.bias",
+      "world_model.grasp_affordance_head.weight",
+      "world_model.grasp_affordance_head.bias",
+      "world_model.support_stability_head.weight",
+      "world_model.support_stability_head.bias",
+      "world_model.persistence_head.weight",
+      "world_model.persistence_head.bias",
+      "world_model.reocclusion_head.weight",
+      "world_model.reocclusion_head.bias",
+      "world_model.disturbance_head.weight",
+      "world_model.disturbance_head.bias",
+      "world_model.uncertainty_head.weight",
+      "world_model.uncertainty_head.bias",
+      "world_model.access_head.weight",
+      "world_model.access_head.bias",
+      "world_model.task_embedding.weight",
+      "world_model.spatial_field_encoder.0.weight",
+      "world_model.spatial_field_encoder.0.bias",
+      "world_model.spatial_field_encoder.2.weight",
+      "world_model.spatial_field_encoder.2.bias",
+      "world_model.spatial_context_proj.0.weight",
+      "world_model.spatial_context_proj.0.bias",
+      "world_model.spatial_context_proj.1.weight",
+      "world_model.spatial_context_proj.1.bias",
+      "world_model.spatial_gate_z.weight",
+      "world_model.spatial_gate_z.bias",
+      "world_model.spatial_gate_r.weight",
+      "world_model.spatial_gate_r.bias",
+      "world_model.spatial_candidate.weight",
+      "world_model.spatial_candidate.bias",
+      "world_model.spatial_summary_proj.0.weight",
+      "world_model.spatial_summary_proj.0.bias",
+      "world_model.spatial_summary_proj.1.weight",
+      "world_model.spatial_summary_proj.1.bias",
+      "world_model.spatial_phase_head.weight",
+      "world_model.spatial_phase_head.bias",
+      "world_model.spatial_support_mode_head.weight",
+      "world_model.spatial_support_mode_head.bias",
+      "world_model.spatial_arm_role_head.weight",
+      "world_model.spatial_arm_role_head.bias",
+      "world_model.spatial_reocclusion_head.weight",
+      "world_model.spatial_reocclusion_head.bias",
+      "world_model.spatial_target_belief_head.weight",
+      "world_model.spatial_target_belief_head.bias",
+      "world_model.spatial_visibility_head.weight",
+      "world_model.spatial_visibility_head.bias",
+      "world_model.spatial_clearance_head.weight",
+      "world_model.spatial_clearance_head.bias",
+      "world_model.spatial_occluder_contact_head.weight",
+      "world_model.spatial_occluder_contact_head.bias",
+      "world_model.spatial_grasp_affordance_head.weight",
+      "world_model.spatial_grasp_affordance_head.bias",
+      "world_model.spatial_support_stability_head.weight",
+      "world_model.spatial_support_stability_head.bias",
+      "world_model.spatial_persistence_head.weight",
+      "world_model.spatial_persistence_head.bias",
+      "world_model.spatial_reocclusion_field_head.weight",
+      "world_model.spatial_reocclusion_field_head.bias",
+      "world_model.spatial_disturbance_head.weight",
+      "world_model.spatial_disturbance_head.bias",
+      "world_model.spatial_uncertainty_head.weight",
+      "world_model.spatial_uncertainty_head.bias",
+      "world_model.spatial_access_head.weight",
+      "world_model.spatial_access_head.bias",
+      "planner.residual.trunk.0.weight",
+      "planner.residual.trunk.0.bias",
+      "planner.residual.trunk.1.weight",
+      "planner.residual.trunk.1.bias",
+      "planner.residual.trunk.3.weight",
+      "planner.residual.trunk.3.bias",
+      "planner.residual.success_head.weight",
+      "planner.residual.success_head.bias",
+      "planner.residual.risk_head.weight",
+      "planner.residual.risk_head.bias",
+      "planner.residual.residual_head.weight",
+      "planner.residual.residual_head.bias"
+    ],
+    "remapped_keys": {},
+    "missing_keys": [
+      "memory.gru.weight_ih_l0",
+      "memory.gru.weight_hh_l0",
+      "memory.gru.bias_ih_l0",
+      "memory.gru.bias_hh_l0",
+      "memory.gru.weight_ih_l1",
+      "memory.gru.weight_hh_l1",
+      "memory.gru.bias_ih_l1",
+      "memory.gru.bias_hh_l1",
+      "memory.token_proj.0.weight",
+      "memory.token_proj.0.bias",
+      "memory.token_proj.1.weight",
+      "memory.token_proj.1.bias",
+      "memory.action_proj.0.weight",
+      "memory.action_proj.0.bias",
+      "memory.action_proj.1.weight",
+      "memory.action_proj.1.bias",
+      "decoder.actor_role_bias",
+      "decoder.revealer_decoder.layers.0.self_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.0.self_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.0.self_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.0.self_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.0.multihead_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.0.multihead_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.0.multihead_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.0.multihead_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.0.linear1.weight",
+      "decoder.revealer_decoder.layers.0.linear1.bias",
+      "decoder.revealer_decoder.layers.0.linear2.weight",
+      "decoder.revealer_decoder.layers.0.linear2.bias",
+      "decoder.revealer_decoder.layers.0.norm1.weight",
+      "decoder.revealer_decoder.layers.0.norm1.bias",
+      "decoder.revealer_decoder.layers.0.norm2.weight",
+      "decoder.revealer_decoder.layers.0.norm2.bias",
+      "decoder.revealer_decoder.layers.0.norm3.weight",
+      "decoder.revealer_decoder.layers.0.norm3.bias",
+      "decoder.revealer_decoder.layers.1.self_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.1.self_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.1.self_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.1.self_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.1.multihead_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.1.multihead_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.1.multihead_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.1.multihead_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.1.linear1.weight",
+      "decoder.revealer_decoder.layers.1.linear1.bias",
+      "decoder.revealer_decoder.layers.1.linear2.weight",
+      "decoder.revealer_decoder.layers.1.linear2.bias",
+      "decoder.revealer_decoder.layers.1.norm1.weight",
+      "decoder.revealer_decoder.layers.1.norm1.bias",
+      "decoder.revealer_decoder.layers.1.norm2.weight",
+      "decoder.revealer_decoder.layers.1.norm2.bias",
+      "decoder.revealer_decoder.layers.1.norm3.weight",
+      "decoder.revealer_decoder.layers.1.norm3.bias",
+      "decoder.revealer_decoder.layers.2.self_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.2.self_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.2.self_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.2.self_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.2.multihead_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.2.multihead_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.2.multihead_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.2.multihead_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.2.linear1.weight",
+      "decoder.revealer_decoder.layers.2.linear1.bias",
+      "decoder.revealer_decoder.layers.2.linear2.weight",
+      "decoder.revealer_decoder.layers.2.linear2.bias",
+      "decoder.revealer_decoder.layers.2.norm1.weight",
+      "decoder.revealer_decoder.layers.2.norm1.bias",
+      "decoder.revealer_decoder.layers.2.norm2.weight",
+      "decoder.revealer_decoder.layers.2.norm2.bias",
+      "decoder.revealer_decoder.layers.2.norm3.weight",
+      "decoder.revealer_decoder.layers.2.norm3.bias",
+      "decoder.revealer_decoder.layers.3.self_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.3.self_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.3.self_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.3.self_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.3.multihead_attn.in_proj_weight",
+      "decoder.revealer_decoder.layers.3.multihead_attn.in_proj_bias",
+      "decoder.revealer_decoder.layers.3.multihead_attn.out_proj.weight",
+      "decoder.revealer_decoder.layers.3.multihead_attn.out_proj.bias",
+      "decoder.revealer_decoder.layers.3.linear1.weight",
+      "decoder.revealer_decoder.layers.3.linear1.bias",
+      "decoder.revealer_decoder.layers.3.linear2.weight",
+      "decoder.revealer_decoder.layers.3.linear2.bias",
+      "decoder.revealer_decoder.layers.3.norm1.weight",
+      "decoder.revealer_decoder.layers.3.norm1.bias",
+      "decoder.revealer_decoder.layers.3.norm2.weight",
+      "decoder.revealer_decoder.layers.3.norm2.bias",
+      "decoder.revealer_decoder.layers.3.norm3.weight",
+      "decoder.revealer_decoder.layers.3.norm3.bias",
+      "decoder.actor_decoder.layers.0.self_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.0.self_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.0.self_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.0.self_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.0.multihead_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.0.multihead_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.0.multihead_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.0.multihead_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.0.linear1.weight",
+      "decoder.actor_decoder.layers.0.linear1.bias",
+      "decoder.actor_decoder.layers.0.linear2.weight",
+      "decoder.actor_decoder.layers.0.linear2.bias",
+      "decoder.actor_decoder.layers.0.norm1.weight",
+      "decoder.actor_decoder.layers.0.norm1.bias",
+      "decoder.actor_decoder.layers.0.norm2.weight",
+      "decoder.actor_decoder.layers.0.norm2.bias",
+      "decoder.actor_decoder.layers.0.norm3.weight",
+      "decoder.actor_decoder.layers.0.norm3.bias",
+      "decoder.actor_decoder.layers.1.self_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.1.self_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.1.self_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.1.self_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.1.multihead_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.1.multihead_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.1.multihead_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.1.multihead_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.1.linear1.weight",
+      "decoder.actor_decoder.layers.1.linear1.bias",
+      "decoder.actor_decoder.layers.1.linear2.weight",
+      "decoder.actor_decoder.layers.1.linear2.bias",
+      "decoder.actor_decoder.layers.1.norm1.weight",
+      "decoder.actor_decoder.layers.1.norm1.bias",
+      "decoder.actor_decoder.layers.1.norm2.weight",
+      "decoder.actor_decoder.layers.1.norm2.bias",
+      "decoder.actor_decoder.layers.1.norm3.weight",
+      "decoder.actor_decoder.layers.1.norm3.bias",
+      "decoder.actor_decoder.layers.2.self_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.2.self_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.2.self_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.2.self_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.2.multihead_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.2.multihead_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.2.multihead_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.2.multihead_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.2.linear1.weight",
+      "decoder.actor_decoder.layers.2.linear1.bias",
+      "decoder.actor_decoder.layers.2.linear2.weight",
+      "decoder.actor_decoder.layers.2.linear2.bias",
+      "decoder.actor_decoder.layers.2.norm1.weight",
+      "decoder.actor_decoder.layers.2.norm1.bias",
+      "decoder.actor_decoder.layers.2.norm2.weight",
+      "decoder.actor_decoder.layers.2.norm2.bias",
+      "decoder.actor_decoder.layers.2.norm3.weight",
+      "decoder.actor_decoder.layers.2.norm3.bias",
+      "decoder.actor_decoder.layers.3.self_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.3.self_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.3.self_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.3.self_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.3.multihead_attn.in_proj_weight",
+      "decoder.actor_decoder.layers.3.multihead_attn.in_proj_bias",
+      "decoder.actor_decoder.layers.3.multihead_attn.out_proj.weight",
+      "decoder.actor_decoder.layers.3.multihead_attn.out_proj.bias",
+      "decoder.actor_decoder.layers.3.linear1.weight",
+      "decoder.actor_decoder.layers.3.linear1.bias",
+      "decoder.actor_decoder.layers.3.linear2.weight",
+      "decoder.actor_decoder.layers.3.linear2.bias",
+      "decoder.actor_decoder.layers.3.norm1.weight",
+      "decoder.actor_decoder.layers.3.norm1.bias",
+      "decoder.actor_decoder.layers.3.norm2.weight",
+      "decoder.actor_decoder.layers.3.norm2.bias",
+      "decoder.actor_decoder.layers.3.norm3.weight",
+      "decoder.actor_decoder.layers.3.norm3.bias",
+      "decoder.revealer_mean.weight",
+      "decoder.revealer_mean.bias",
+      "decoder.revealer_log_std.weight",
+      "decoder.revealer_log_std.bias",
+      "decoder.actor_mean.weight",
+      "decoder.actor_mean.bias",
+      "decoder.actor_log_std.weight",
+      "decoder.actor_log_std.bias",
+      "decoder.proposal_score.0.weight",
+      "decoder.proposal_score.0.bias",
+      "decoder.proposal_score.1.weight",
+      "decoder.proposal_score.1.bias"
+    ],
+    "unexpected_keys": []
+  },
+  "trainable_parameter_names": [
+    "fusion.camera_embedding.weight",
+    "fusion.cross_view_transformer.layers.0.self_attn.in_proj_weight",
+    "fusion.cross_view_transformer.layers.0.self_attn.in_proj_bias",
+    "fusion.cross_view_transformer.layers.0.self_attn.out_proj.weight",
+    "fusion.cross_view_transformer.layers.0.self_attn.out_proj.bias",
+    "fusion.cross_view_transformer.layers.0.linear1.weight",
+    "fusion.cross_view_transformer.layers.0.linear1.bias",
+    "fusion.cross_view_transformer.layers.0.linear2.weight",
+    "fusion.cross_view_transformer.layers.0.linear2.bias",
+    "fusion.cross_view_transformer.layers.0.norm1.weight",
+    "fusion.cross_view_transformer.layers.0.norm1.bias",
+    "fusion.cross_view_transformer.layers.0.norm2.weight",
+    "fusion.cross_view_transformer.layers.0.norm2.bias",
+    "fusion.cross_view_transformer.layers.1.self_attn.in_proj_weight",
+    "fusion.cross_view_transformer.layers.1.self_attn.in_proj_bias",
+    "fusion.cross_view_transformer.layers.1.self_attn.out_proj.weight",
+    "fusion.cross_view_transformer.layers.1.self_attn.out_proj.bias",
+    "fusion.cross_view_transformer.layers.1.linear1.weight",
+    "fusion.cross_view_transformer.layers.1.linear1.bias",
+    "fusion.cross_view_transformer.layers.1.linear2.weight",
+    "fusion.cross_view_transformer.layers.1.linear2.bias",
+    "fusion.cross_view_transformer.layers.1.norm1.weight",
+    "fusion.cross_view_transformer.layers.1.norm1.bias",
+    "fusion.cross_view_transformer.layers.1.norm2.weight",
+    "fusion.cross_view_transformer.layers.1.norm2.bias",
+    "fusion.cross_view_transformer.layers.2.self_attn.in_proj_weight",
+    "fusion.cross_view_transformer.layers.2.self_attn.in_proj_bias",
+    "fusion.cross_view_transformer.layers.2.self_attn.out_proj.weight",
+    "fusion.cross_view_transformer.layers.2.self_attn.out_proj.bias",
+    "fusion.cross_view_transformer.layers.2.linear1.weight",
+    "fusion.cross_view_transformer.layers.2.linear1.bias",
+    "fusion.cross_view_transformer.layers.2.linear2.weight",
+    "fusion.cross_view_transformer.layers.2.linear2.bias",
+    "fusion.cross_view_transformer.layers.2.norm1.weight",
+    "fusion.cross_view_transformer.layers.2.norm1.bias",
+    "fusion.cross_view_transformer.layers.2.norm2.weight",
+    "fusion.cross_view_transformer.layers.2.norm2.bias",
+    "fusion.cross_view_transformer.layers.3.self_attn.in_proj_weight",
+    "fusion.cross_view_transformer.layers.3.self_attn.in_proj_bias",
+    "fusion.cross_view_transformer.layers.3.self_attn.out_proj.weight",
+    "fusion.cross_view_transformer.layers.3.self_attn.out_proj.bias",
+    "fusion.cross_view_transformer.layers.3.linear1.weight",
+    "fusion.cross_view_transformer.layers.3.linear1.bias",
+    "fusion.cross_view_transformer.layers.3.linear2.weight",
+    "fusion.cross_view_transformer.layers.3.linear2.bias",
+    "fusion.cross_view_transformer.layers.3.norm1.weight",
+    "fusion.cross_view_transformer.layers.3.norm1.bias",
+    "fusion.cross_view_transformer.layers.3.norm2.weight",
+    "fusion.cross_view_transformer.layers.3.norm2.bias",
+    "fusion.geometry_fusion.attn.in_proj_weight",
+    "fusion.geometry_fusion.attn.in_proj_bias",
+    "fusion.geometry_fusion.attn.out_proj.weight",
+    "fusion.geometry_fusion.attn.out_proj.bias",
+    "fusion.geometry_fusion.gate.0.weight",
+    "fusion.geometry_fusion.gate.0.bias",
+    "fusion.geometry_fusion.gate.1.weight",
+    "fusion.geometry_fusion.gate.1.bias",
+    "fusion.geometry_fusion.gate.3.weight",
+    "fusion.geometry_fusion.gate.3.bias",
+    "fusion.geometry_fusion.out.0.weight",
+    "fusion.geometry_fusion.out.0.bias",
+    "fusion.geometry_fusion.out.1.weight",
+    "fusion.geometry_fusion.out.1.bias",
+    "fusion.proprio_adapter.0.weight",
+    "fusion.proprio_adapter.0.bias",
+    "fusion.proprio_adapter.1.weight",
+    "fusion.proprio_adapter.1.bias",
+    "memory.gru.weight_ih_l0",
+    "memory.gru.weight_hh_l0",
+    "memory.gru.bias_ih_l0",
+    "memory.gru.bias_hh_l0",
+    "memory.gru.weight_ih_l1",
+    "memory.gru.weight_hh_l1",
+    "memory.gru.bias_ih_l1",
+    "memory.gru.bias_hh_l1",
+    "memory.token_proj.0.weight",
+    "memory.token_proj.0.bias",
+    "memory.token_proj.1.weight",
+    "memory.token_proj.1.bias",
+    "memory.action_proj.0.weight",
+    "memory.action_proj.0.bias",
+    "memory.action_proj.1.weight",
+    "memory.action_proj.1.bias",
+    "memory.uncertainty_head.0.weight",
+    "memory.uncertainty_head.0.bias",
+    "memory.uncertainty_head.1.weight",
+    "memory.uncertainty_head.1.bias",
+    "decoder.actor_role_bias",
+    "decoder.revealer_decoder.layers.0.self_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.0.self_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.0.self_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.0.self_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.0.multihead_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.0.multihead_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.0.multihead_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.0.multihead_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.0.linear1.weight",
+    "decoder.revealer_decoder.layers.0.linear1.bias",
+    "decoder.revealer_decoder.layers.0.linear2.weight",
+    "decoder.revealer_decoder.layers.0.linear2.bias",
+    "decoder.revealer_decoder.layers.0.norm1.weight",
+    "decoder.revealer_decoder.layers.0.norm1.bias",
+    "decoder.revealer_decoder.layers.0.norm2.weight",
+    "decoder.revealer_decoder.layers.0.norm2.bias",
+    "decoder.revealer_decoder.layers.0.norm3.weight",
+    "decoder.revealer_decoder.layers.0.norm3.bias",
+    "decoder.revealer_decoder.layers.1.self_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.1.self_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.1.self_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.1.self_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.1.multihead_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.1.multihead_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.1.multihead_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.1.multihead_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.1.linear1.weight",
+    "decoder.revealer_decoder.layers.1.linear1.bias",
+    "decoder.revealer_decoder.layers.1.linear2.weight",
+    "decoder.revealer_decoder.layers.1.linear2.bias",
+    "decoder.revealer_decoder.layers.1.norm1.weight",
+    "decoder.revealer_decoder.layers.1.norm1.bias",
+    "decoder.revealer_decoder.layers.1.norm2.weight",
+    "decoder.revealer_decoder.layers.1.norm2.bias",
+    "decoder.revealer_decoder.layers.1.norm3.weight",
+    "decoder.revealer_decoder.layers.1.norm3.bias",
+    "decoder.revealer_decoder.layers.2.self_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.2.self_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.2.self_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.2.self_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.2.multihead_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.2.multihead_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.2.multihead_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.2.multihead_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.2.linear1.weight",
+    "decoder.revealer_decoder.layers.2.linear1.bias",
+    "decoder.revealer_decoder.layers.2.linear2.weight",
+    "decoder.revealer_decoder.layers.2.linear2.bias",
+    "decoder.revealer_decoder.layers.2.norm1.weight",
+    "decoder.revealer_decoder.layers.2.norm1.bias",
+    "decoder.revealer_decoder.layers.2.norm2.weight",
+    "decoder.revealer_decoder.layers.2.norm2.bias",
+    "decoder.revealer_decoder.layers.2.norm3.weight",
+    "decoder.revealer_decoder.layers.2.norm3.bias",
+    "decoder.revealer_decoder.layers.3.self_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.3.self_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.3.self_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.3.self_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.3.multihead_attn.in_proj_weight",
+    "decoder.revealer_decoder.layers.3.multihead_attn.in_proj_bias",
+    "decoder.revealer_decoder.layers.3.multihead_attn.out_proj.weight",
+    "decoder.revealer_decoder.layers.3.multihead_attn.out_proj.bias",
+    "decoder.revealer_decoder.layers.3.linear1.weight",
+    "decoder.revealer_decoder.layers.3.linear1.bias",
+    "decoder.revealer_decoder.layers.3.linear2.weight",
+    "decoder.revealer_decoder.layers.3.linear2.bias",
+    "decoder.revealer_decoder.layers.3.norm1.weight",
+    "decoder.revealer_decoder.layers.3.norm1.bias",
+    "decoder.revealer_decoder.layers.3.norm2.weight",
+    "decoder.revealer_decoder.layers.3.norm2.bias",
+    "decoder.revealer_decoder.layers.3.norm3.weight",
+    "decoder.revealer_decoder.layers.3.norm3.bias",
+    "decoder.actor_decoder.layers.0.self_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.0.self_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.0.self_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.0.self_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.0.multihead_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.0.multihead_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.0.multihead_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.0.multihead_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.0.linear1.weight",
+    "decoder.actor_decoder.layers.0.linear1.bias",
+    "decoder.actor_decoder.layers.0.linear2.weight",
+    "decoder.actor_decoder.layers.0.linear2.bias",
+    "decoder.actor_decoder.layers.0.norm1.weight",
+    "decoder.actor_decoder.layers.0.norm1.bias",
+    "decoder.actor_decoder.layers.0.norm2.weight",
+    "decoder.actor_decoder.layers.0.norm2.bias",
+    "decoder.actor_decoder.layers.0.norm3.weight",
+    "decoder.actor_decoder.layers.0.norm3.bias",
+    "decoder.actor_decoder.layers.1.self_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.1.self_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.1.self_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.1.self_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.1.multihead_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.1.multihead_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.1.multihead_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.1.multihead_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.1.linear1.weight",
+    "decoder.actor_decoder.layers.1.linear1.bias",
+    "decoder.actor_decoder.layers.1.linear2.weight",
+    "decoder.actor_decoder.layers.1.linear2.bias",
+    "decoder.actor_decoder.layers.1.norm1.weight",
+    "decoder.actor_decoder.layers.1.norm1.bias",
+    "decoder.actor_decoder.layers.1.norm2.weight",
+    "decoder.actor_decoder.layers.1.norm2.bias",
+    "decoder.actor_decoder.layers.1.norm3.weight",
+    "decoder.actor_decoder.layers.1.norm3.bias",
+    "decoder.actor_decoder.layers.2.self_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.2.self_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.2.self_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.2.self_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.2.multihead_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.2.multihead_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.2.multihead_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.2.multihead_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.2.linear1.weight",
+    "decoder.actor_decoder.layers.2.linear1.bias",
+    "decoder.actor_decoder.layers.2.linear2.weight",
+    "decoder.actor_decoder.layers.2.linear2.bias",
+    "decoder.actor_decoder.layers.2.norm1.weight",
+    "decoder.actor_decoder.layers.2.norm1.bias",
+    "decoder.actor_decoder.layers.2.norm2.weight",
+    "decoder.actor_decoder.layers.2.norm2.bias",
+    "decoder.actor_decoder.layers.2.norm3.weight",
+    "decoder.actor_decoder.layers.2.norm3.bias",
+    "decoder.actor_decoder.layers.3.self_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.3.self_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.3.self_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.3.self_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.3.multihead_attn.in_proj_weight",
+    "decoder.actor_decoder.layers.3.multihead_attn.in_proj_bias",
+    "decoder.actor_decoder.layers.3.multihead_attn.out_proj.weight",
+    "decoder.actor_decoder.layers.3.multihead_attn.out_proj.bias",
+    "decoder.actor_decoder.layers.3.linear1.weight",
+    "decoder.actor_decoder.layers.3.linear1.bias",
+    "decoder.actor_decoder.layers.3.linear2.weight",
+    "decoder.actor_decoder.layers.3.linear2.bias",
+    "decoder.actor_decoder.layers.3.norm1.weight",
+    "decoder.actor_decoder.layers.3.norm1.bias",
+    "decoder.actor_decoder.layers.3.norm2.weight",
+    "decoder.actor_decoder.layers.3.norm2.bias",
+    "decoder.actor_decoder.layers.3.norm3.weight",
+    "decoder.actor_decoder.layers.3.norm3.bias",
+    "decoder.query_embed.weight",
+    "decoder.revealer_mean.weight",
+    "decoder.revealer_mean.bias",
+    "decoder.revealer_log_std.weight",
+    "decoder.revealer_log_std.bias",
+    "decoder.actor_mean.weight",
+    "decoder.actor_mean.bias",
+    "decoder.actor_log_std.weight",
+    "decoder.actor_log_std.bias",
+    "decoder.coordination.0.weight",
+    "decoder.coordination.0.bias",
+    "decoder.coordination.1.weight",
+    "decoder.coordination.1.bias",
+    "decoder.coordination.3.weight",
+    "decoder.coordination.3.bias",
+    "decoder.proposal_score.0.weight",
+    "decoder.proposal_score.0.bias",
+    "decoder.proposal_score.1.weight",
+    "decoder.proposal_score.1.bias"
+  ],
+  "best_val_total": 0.33879721803324564,
+  "history": [
+    {
+      "epoch": 0,
+      "train": {
+        "action": 0.6207457804366162,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.210943855737385,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.7382262051105499,
+        "role_swap_consistency": 0.0,
+        "total": 0.8909744981088137,
+        "transition": 0.0,
+        "world_model": 0.0
+      },
+      "val": {
+        "action": 0.19714589416980743,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.1202252422060286,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.6674523949623108,
+        "role_swap_consistency": 0.0,
+        "total": 0.44527397411210196,
+        "transition": 0.0,
+        "world_model": 0.0
+      }
+    },
+    {
+      "epoch": 1,
+      "train": {
+        "action": 0.17574271442074524,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.2226158066799766,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.7345987056431017,
+        "role_swap_consistency": 0.0,
+        "total": 0.44728693836613703,
+        "transition": 0.0,
+        "world_model": 0.0
+      },
+      "val": {
+        "action": 0.1045121390904699,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.2785813467843192,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.7802676217896598,
+        "role_swap_consistency": 0.0,
+        "total": 0.3899314616407667,
+        "transition": 0.0,
+        "world_model": 0.0
+      }
+    },
+    {
+      "epoch": 2,
+      "train": {
+        "action": 0.18294762859219,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.2462087493193776,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.7633898587603318,
+        "role_swap_consistency": 0.0,
+        "total": 0.46148573254284103,
+        "transition": 0.0,
+        "world_model": 0.0
+      },
+      "val": {
+        "action": 0.10954179934092931,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.1277755498886108,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.6412685768944877,
+        "role_swap_consistency": 0.0,
+        "total": 0.355660366160529,
+        "transition": 0.0,
+        "world_model": 0.0
+      }
+    },
+    {
+      "epoch": 3,
+      "train": {
+        "action": 0.175935955031922,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.230568854432357,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.7617053389549255,
+        "role_swap_consistency": 0.0,
+        "total": 0.4519259302239669,
+        "transition": 0.0,
+        "world_model": 0.0
+      },
+      "val": {
+        "action": 0.0885860047170094,
+        "distillation": 0.0,
+        "gate": 0.0,
+        "planner_ranking": 0.0,
+        "planner_risk": 0.0,
+        "planner_success": 0.0,
+        "proposal_diversity": 0.0,
+        "proposal_mode": 0.0,
+        "proposal_ranking": 1.1679236718586512,
+        "proposal_reconstruction": 0.0,
+        "proposal_success": 0.6251888232571738,
+        "role_swap_consistency": 0.0,
+        "total": 0.33879721803324564,
+        "transition": 0.0,
+        "world_model": 0.0
+      }
+    }
+  ],
+  "train_spec": {
+    "track_id": "bag_track",
+    "suite": "maniskill3",
+    "benchmark_task": "PutEggplantInBasketRetrievalProxy-v1",
+    "model_variant": "trunk_only_ft",
+    "seed": 23,
+    "train_demos": 32,
+    "val_demos": 8,
+    "init_checkpoint_group": "/workspace/workspace/VLAarchtests2/VLAarchtests/artifacts/outputs/r3d_handoff_phase/proxy_interaction_r3d_stage3_clip_rgbd_handoff_compact_phase_seed17/checkpoint_best.pt",
+    "optimizer": "adamw",
+    "learning_rate": 0.0001,
+    "lr_schedule": "constant",
+    "batch_size": 4,
+    "augmentations": "none",
+    "early_stopping_metric": "val_total",
+    "max_gradient_steps": 114,
+    "unfreeze_scope": "fusion_memory_decoder",
+    "dataset_split_id": "bag_bridge_smoke_v1_dataset_seed17",
+    "same_data_policy": true,
+    "same_init_policy": true
+  }
+}

outputs/maniskill_cloth_bridge_smoke_v1/adapter_active_ft_seed17/checkpoint_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9927c95e5aea8dff34444bf04db80491212579a22df8a56e568851cdbcba243
+size 878859044

outputs/maniskill_cloth_bridge_smoke_v1/adapter_active_ft_seed23/checkpoint_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee11fa2e4f3d467201da345a1da4752a563c87337c7c6183ede39c3246319302
+size 878859044

outputs/maniskill_cloth_bridge_smoke_v1/adapter_active_ft_seed29/checkpoint_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ccd061ab8d90059a771b4b8079f0ccb9884b1b30a95a895b7276331c35a0f016
+size 878859044