Auto-sync: 2026-06-30 09:56:41 (part 2)

results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types6_prepend_margin0p00_summary.json

@@ -0,0 +1,427 @@
+{
+  "run_root": "/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs",
+  "objective": "near_miss_policy_bc5_typedprop_p2",
+  "out_name": "policy_rollout_proposal_lattice_types6_prepend_margin0p00.json",
+  "num_completed": 3,
+  "baseline_h4_policy_success": 0.2967,
+  "baseline_h16_rank_checkpoint_success": 0.29739130434782607,
+  "mean_success": 0.3130434782608696,
+  "std_success": 0.0381421081729762,
+  "mean_progress": 0.5402942887342373,
+  "mean_action_mse_to_best": 0.4803936158428205,
+  "gain_vs_h4": 0.01634347826086957,
+  "gain_vs_h16_rank_checkpoint": 0.01565217391304352,
+  "rows": [
+    {
+      "seed": 0,
+      "path": "/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs/near_miss_policy_bc5_typedprop_p2/seed_0/policy_rollout_proposal_lattice_types6_prepend_margin0p00.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 7,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "selection_margin": 0.0,
+      "prepend_policy_candidate": true,
+      "field_optim_steps": 0,
+      "field_optim_step_size": 0.0,
+      "field_optim_trust_radius": 0.0,
+      "field_optim_l2_penalty": 0.0,
+      "retrieval_neighbors": 0,
+      "retrieval_metric": "none",
+      "retrieval_type_min_success": 0.0,
+      "retrieval_type_success_bonus_scale": 0.0,
+      "retrieval_residual_consensus_penalty_scale": 0.0,
+      "retrieval_residual_min_source_progress": 0.0,
+      "retrieval_residual_min_source_advantage": -1000000000.0,
+      "retrieval_residual_source_progress_bonus_scale": 0.0,
+      "retrieval_residual_source_score_bonus_scale": 0.0,
+      "retrieval_residual_source_advantage_bonus_scale": 0.0,
+      "retrieval_residual_composite_l2_penalty_scale": 0.0,
+      "retrieval_residual_scale": 0.0,
+      "retrieval_residual_scales": [],
+      "retrieval_residual_anchor": "none",
+      "retrieval_residual_direction": "none",
+      "retrieval_residual_reduce": "none",
+      "retrieval_residual_challenger_types": [],
+      "retrieval_residual_challenger_scales": [],
+      "retrieval_residual_challenger_margin": 0.0,
+      "lattice_exclude_types": [],
+      "candidate_type_bonuses": {},
+      "candidate_type_bonus_components": false,
+      "candidate_oracle_rollouts": 0,
+      "candidate_oracle_unique_tolerance": 1e-06,
+      "candidate_oracle_success_rate": null,
+      "candidate_oracle_progress": null,
+      "candidate_oracle_score": null,
+      "candidate_oracle_regret": null,
+      "candidate_oracle_score_gain_over_selected": null,
+      "candidate_oracle_unique_count": null,
+      "candidate_oracle_improvement_rate": null,
+      "candidate_oracle_selected_branch_success_rate": null,
+      "candidate_oracle_selected_branch_progress": null,
+      "candidate_oracle_type_counts": {},
+      "candidate_oracle_best_branch_rank": null,
+      "candidate_oracle_best_branch_rank_counts": {},
+      "candidate_oracle_branch_success_rates": [],
+      "candidate_oracle_branch_progress": [],
+      "candidate_oracle_branch_score_gains_over_selected": [],
+      "selected_residual_scale_counts": {},
+      "policy_rollout_success_rate": 0.26956521739130435,
+      "policy_rollout_progress": 0.5106911778385225,
+      "oracle_success_rate": 0.8573913043478261,
+      "action_mse_to_best": 0.4307826819503174,
+      "best_policy_val": {
+        "bc_loss": 0.1345199979841709,
+        "field_effect_loss": 0.009223711728635762,
+        "field_potential_loss": 0.2614195512400733,
+        "field_preference_loss": 0.5129645930396186,
+        "lattice_edges": 3833.3333333333335,
+        "progress_mae": 0.18681737035512924,
+        "proposal_loss": 0.15201550225416818,
+        "rank_acc": 0.8316339155038198,
+        "rank_loss": 0.5129594521390067,
+        "regret_mae": 0.3556148376729753,
+        "success_accuracy": 0.8784897228082021,
+        "total_loss": 1.8417018850644429
+      },
+      "per_task": {
+        "LiftPegUpright-v1": {
+          "action_mse_to_best": 0.3190493986644235,
+          "expert_success_rate": 0.8865979381443299,
+          "num_groups": 97,
+          "oracle_success_rate": 0.9278350515463918,
+          "policy_expert_regret": 1.3113837868896956,
+          "policy_oracle_regret": 1.364168087231744,
+          "policy_rollout_progress": 0.4357556242918231,
+          "policy_rollout_success_rate": 0.10309278350515463,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.3754520480118887,
+          "expert_success_rate": 0.9375,
+          "num_groups": 208,
+          "oracle_success_rate": 0.9471153846153846,
+          "policy_expert_regret": 1.147191958806406,
+          "policy_oracle_regret": 1.1472131935342287,
+          "policy_rollout_progress": 0.554053240795978,
+          "policy_rollout_success_rate": 0.23076923076923078,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.6732666387083613,
+          "expert_success_rate": 0.19480519480519481,
+          "num_groups": 77,
+          "oracle_success_rate": 0.36363636363636365,
+          "policy_expert_regret": 0.29871676349407666,
+          "policy_oracle_regret": 0.5623729637303909,
+          "policy_rollout_progress": 0.2738164303364692,
+          "policy_rollout_success_rate": 0.1038961038961039,
+          "restore_max_error": 2.3096799850463867e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.40638454278257297,
+          "expert_success_rate": 0.8617021276595744,
+          "num_groups": 94,
+          "oracle_success_rate": 0.9893617021276596,
+          "policy_expert_regret": 0.3492174619372855,
+          "policy_oracle_regret": 0.4675355685201097,
+          "policy_rollout_progress": 0.7681559687282177,
+          "policy_rollout_success_rate": 0.7446808510638298,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.4910762248727766,
+          "expert_success_rate": 0.6767676767676768,
+          "num_groups": 99,
+          "oracle_success_rate": 0.8585858585858586,
+          "policy_expert_regret": 1.0065984899046445,
+          "policy_oracle_regret": 1.2079735342601332,
+          "policy_rollout_progress": 0.4327831456456522,
+          "policy_rollout_success_rate": 0.1919191919191919,
+          "restore_max_error": 2.384185791015625e-07
+        }
+      }
+    },
+    {
+      "seed": 1,
+      "path": "/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs/near_miss_policy_bc5_typedprop_p2/seed_1/policy_rollout_proposal_lattice_types6_prepend_margin0p00.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 7,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "selection_margin": 0.0,
+      "prepend_policy_candidate": true,
+      "field_optim_steps": 0,
+      "field_optim_step_size": 0.0,
+      "field_optim_trust_radius": 0.0,
+      "field_optim_l2_penalty": 0.0,
+      "retrieval_neighbors": 0,
+      "retrieval_metric": "none",
+      "retrieval_type_min_success": 0.0,
+      "retrieval_type_success_bonus_scale": 0.0,
+      "retrieval_residual_consensus_penalty_scale": 0.0,
+      "retrieval_residual_min_source_progress": 0.0,
+      "retrieval_residual_min_source_advantage": -1000000000.0,
+      "retrieval_residual_source_progress_bonus_scale": 0.0,
+      "retrieval_residual_source_score_bonus_scale": 0.0,
+      "retrieval_residual_source_advantage_bonus_scale": 0.0,
+      "retrieval_residual_composite_l2_penalty_scale": 0.0,
+      "retrieval_residual_scale": 0.0,
+      "retrieval_residual_scales": [],
+      "retrieval_residual_anchor": "none",
+      "retrieval_residual_direction": "none",
+      "retrieval_residual_reduce": "none",
+      "retrieval_residual_challenger_types": [],
+      "retrieval_residual_challenger_scales": [],
+      "retrieval_residual_challenger_margin": 0.0,
+      "lattice_exclude_types": [],
+      "candidate_type_bonuses": {},
+      "candidate_type_bonus_components": false,
+      "candidate_oracle_rollouts": 0,
+      "candidate_oracle_unique_tolerance": 1e-06,
+      "candidate_oracle_success_rate": null,
+      "candidate_oracle_progress": null,
+      "candidate_oracle_score": null,
+      "candidate_oracle_regret": null,
+      "candidate_oracle_score_gain_over_selected": null,
+      "candidate_oracle_unique_count": null,
+      "candidate_oracle_improvement_rate": null,
+      "candidate_oracle_selected_branch_success_rate": null,
+      "candidate_oracle_selected_branch_progress": null,
+      "candidate_oracle_type_counts": {},
+      "candidate_oracle_best_branch_rank": null,
+      "candidate_oracle_best_branch_rank_counts": {},
+      "candidate_oracle_branch_success_rates": [],
+      "candidate_oracle_branch_progress": [],
+      "candidate_oracle_branch_score_gains_over_selected": [],
+      "selected_residual_scale_counts": {},
+      "policy_rollout_success_rate": 0.3408695652173913,
+      "policy_rollout_progress": 0.5600207342295502,
+      "oracle_success_rate": 0.8695652173913043,
+      "action_mse_to_best": 0.47603451989216566,
+      "best_policy_val": {
+        "bc_loss": 0.12785478805502257,
+        "field_effect_loss": 0.010064840265032318,
+        "field_potential_loss": 0.29153423342439866,
+        "field_preference_loss": 0.5177145335409377,
+        "lattice_edges": 3833.3333333333335,
+        "progress_mae": 0.19837388147910437,
+        "proposal_loss": 0.14544310627712143,
+        "rank_acc": 0.8285107778178321,
+        "rank_loss": 0.5177104257875018,
+        "regret_mae": 0.37083542181385887,
+        "success_accuracy": 0.8705372114976248,
+        "total_loss": 1.8337699174880981
+      },
+      "per_task": {
+        "LiftPegUpright-v1": {
+          "action_mse_to_best": 0.3306756934884573,
+          "expert_success_rate": 0.8584070796460177,
+          "num_groups": 113,
+          "oracle_success_rate": 0.9380530973451328,
+          "policy_expert_regret": 0.9908357312721489,
+          "policy_oracle_regret": 1.0945943040130413,
+          "policy_rollout_progress": 0.6228555224638069,
+          "policy_rollout_success_rate": 0.20353982300884957,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.4901451797730102,
+          "expert_success_rate": 0.9402173913043478,
+          "num_groups": 184,
+          "oracle_success_rate": 0.9456521739130435,
+          "policy_expert_regret": 1.0306415203076145,
+          "policy_oracle_regret": 1.0306655247825447,
+          "policy_rollout_progress": 0.5961379015324105,
+          "policy_rollout_success_rate": 0.30434782608695654,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.6031585289533005,
+          "expert_success_rate": 0.25,
+          "num_groups": 76,
+          "oracle_success_rate": 0.40789473684210525,
+          "policy_expert_regret": 0.287915515027156,
+          "policy_oracle_regret": 0.4959764588428171,
+          "policy_rollout_progress": 0.3087349236011505,
+          "policy_rollout_success_rate": 0.17105263157894737,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.4626710661896714,
+          "expert_success_rate": 0.8198198198198198,
+          "num_groups": 111,
+          "oracle_success_rate": 1.0,
+          "policy_expert_regret": 0.4856817053781973,
+          "policy_oracle_regret": 0.5529967908923691,
+          "policy_rollout_progress": 0.7352914973959193,
+          "policy_rollout_success_rate": 0.7117117117117117,
+          "restore_max_error": 2.3795291781425476e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.538134570857817,
+          "expert_success_rate": 0.6923076923076923,
+          "num_groups": 91,
+          "oracle_success_rate": 0.8571428571428571,
+          "policy_expert_regret": 1.030504544848924,
+          "policy_oracle_regret": 1.1395653635263443,
+          "policy_rollout_progress": 0.4050402621646504,
+          "policy_rollout_success_rate": 0.27472527472527475,
+          "restore_max_error": 1.9371509552001953e-07
+        }
+      }
+    },
+    {
+      "seed": 2,
+      "path": "/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs/near_miss_policy_bc5_typedprop_p2/seed_2/policy_rollout_proposal_lattice_types6_prepend_margin0p00.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 7,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "selection_margin": 0.0,
+      "prepend_policy_candidate": true,
+      "field_optim_steps": 0,
+      "field_optim_step_size": 0.0,
+      "field_optim_trust_radius": 0.0,
+      "field_optim_l2_penalty": 0.0,
+      "retrieval_neighbors": 0,
+      "retrieval_metric": "none",
+      "retrieval_type_min_success": 0.0,
+      "retrieval_type_success_bonus_scale": 0.0,
+      "retrieval_residual_consensus_penalty_scale": 0.0,
+      "retrieval_residual_min_source_progress": 0.0,
+      "retrieval_residual_min_source_advantage": -1000000000.0,
+      "retrieval_residual_source_progress_bonus_scale": 0.0,
+      "retrieval_residual_source_score_bonus_scale": 0.0,
+      "retrieval_residual_source_advantage_bonus_scale": 0.0,
+      "retrieval_residual_composite_l2_penalty_scale": 0.0,
+      "retrieval_residual_scale": 0.0,
+      "retrieval_residual_scales": [],
+      "retrieval_residual_anchor": "none",
+      "retrieval_residual_direction": "none",
+      "retrieval_residual_reduce": "none",
+      "retrieval_residual_challenger_types": [],
+      "retrieval_residual_challenger_scales": [],
+      "retrieval_residual_challenger_margin": 0.0,
+      "lattice_exclude_types": [],
+      "candidate_type_bonuses": {},
+      "candidate_type_bonus_components": false,
+      "candidate_oracle_rollouts": 0,
+      "candidate_oracle_unique_tolerance": 1e-06,
+      "candidate_oracle_success_rate": null,
+      "candidate_oracle_progress": null,
+      "candidate_oracle_score": null,
+      "candidate_oracle_regret": null,
+      "candidate_oracle_score_gain_over_selected": null,
+      "candidate_oracle_unique_count": null,
+      "candidate_oracle_improvement_rate": null,
+      "candidate_oracle_selected_branch_success_rate": null,
+      "candidate_oracle_selected_branch_progress": null,
+      "candidate_oracle_type_counts": {},
+      "candidate_oracle_best_branch_rank": null,
+      "candidate_oracle_best_branch_rank_counts": {},
+      "candidate_oracle_branch_success_rates": [],
+      "candidate_oracle_branch_progress": [],
+      "candidate_oracle_branch_score_gains_over_selected": [],
+      "selected_residual_scale_counts": {},
+      "policy_rollout_success_rate": 0.32869565217391306,
+      "policy_rollout_progress": 0.5501709541346392,
+      "oracle_success_rate": 0.8765217391304347,
+      "action_mse_to_best": 0.5343636456859784,
+      "best_policy_val": {
+        "bc_loss": 0.11429175941480531,
+        "field_effect_loss": 0.009631710747877756,
+        "field_potential_loss": 0.2724720388650894,
+        "field_preference_loss": 0.5087620665629705,
+        "lattice_edges": 3833.3333333333335,
+        "progress_mae": 0.19653971824381086,
+        "proposal_loss": 0.13279944616887304,
+        "rank_acc": 0.8339865836832259,
+        "rank_loss": 0.508702975180414,
+        "regret_mae": 0.3901382436354955,
+        "success_accuracy": 0.870691223276986,
+        "total_loss": 1.710942678981357
+      },
+      "per_task": {
+        "LiftPegUpright-v1": {
+          "action_mse_to_best": 0.3514371175309255,
+          "expert_success_rate": 0.8229166666666666,
+          "num_groups": 96,
+          "oracle_success_rate": 0.9270833333333334,
+          "policy_expert_regret": 0.8929005730897188,
+          "policy_oracle_regret": 0.9729711189866066,
+          "policy_rollout_progress": 0.6689869376520315,
+          "policy_rollout_success_rate": 0.2916666666666667,
+          "restore_max_error": 1.955777406692505e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.5641993889917333,
+          "expert_success_rate": 0.9444444444444444,
+          "num_groups": 198,
+          "oracle_success_rate": 0.9595959595959596,
+          "policy_expert_regret": 0.9782833044412499,
+          "policy_oracle_regret": 0.985314491705856,
+          "policy_rollout_progress": 0.6237669653527326,
+          "policy_rollout_success_rate": 0.3333333333333333,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.7903303091556558,
+          "expert_success_rate": 0.24444444444444444,
+          "num_groups": 90,
+          "oracle_success_rate": 0.4666666666666667,
+          "policy_expert_regret": 0.3208057075738907,
+          "policy_oracle_regret": 0.6479979103638066,
+          "policy_rollout_progress": 0.29961686333020526,
+          "policy_rollout_success_rate": 0.14444444444444443,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.3878161942604745,
+          "expert_success_rate": 0.8514851485148515,
+          "num_groups": 101,
+          "oracle_success_rate": 1.0,
+          "policy_expert_regret": 0.5478068379187347,
+          "policy_oracle_regret": 0.6407201064400153,
+          "policy_rollout_progress": 0.6959135569263213,
+          "policy_rollout_success_rate": 0.6633663366336634,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.5723387835754289,
+          "expert_success_rate": 0.7666666666666667,
+          "num_groups": 90,
+          "oracle_success_rate": 0.9111111111111111,
+          "policy_expert_regret": 1.243399785955747,
+          "policy_oracle_regret": 1.3670659497380258,
+          "policy_rollout_progress": 0.34852118359671697,
+          "policy_rollout_success_rate": 0.16666666666666666,
+          "restore_max_error": 2.2351741790771484e-07
+        }
+      }
+    }
+  ]
+}

results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types6_prepend_margin0p00_summary.md

@@ -0,0 +1,19 @@
+# h=16 Best-Policy Checkpoint Rollout
+
+Run root: `/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs`
+Objective: `near_miss_policy_bc5_typedprop_p2`
+Result file: `policy_rollout_proposal_lattice_types6_prepend_margin0p00.json`
+Completed seeds: 3
+Baseline h=4 policy success: 29.67%
+Baseline h=16 rank-checkpoint success: 29.74%
+
+Mean success: 31.30% +/- 3.81%
+Gain vs h=16 rank checkpoint: +1.57%
+Mean progress: 54.03%
+Mean action MSE to best: 0.480
+
+| seed | mode | k | policy cand | retrieval K | retrieval metric | residual anchor | residual direction | residual reduce | min type success | type success bonus | consensus penalty | min source progress | source progress bonus | source score bonus | residual scale | residual scales | margin | sigma | opt steps | trust | success | progress | oracle | candidate oracle | oracle gain | action MSE |
+|---:|---|---:|---|---:|---|---|---|---|---:|---:|---:|---:|---:|---:|---:|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
+| 0 | proposal_lattice | 7 | yes | 0 | none | none | none | none | 0.00 | 0.000 | 0.000 | 0.00 | 0.000 | 0.000 | 0.00 | none | 0.000 | 0.00 | 0 | 0.00 | 26.96% | 51.07% | 85.74% | n/a | n/a | 0.431 |
+| 1 | proposal_lattice | 7 | yes | 0 | none | none | none | none | 0.00 | 0.000 | 0.000 | 0.00 | 0.000 | 0.000 | 0.00 | none | 0.000 | 0.00 | 0 | 0.00 | 34.09% | 56.00% | 86.96% | n/a | n/a | 0.476 |
+| 2 | proposal_lattice | 7 | yes | 0 | none | none | none | none | 0.00 | 0.000 | 0.000 | 0.00 | 0.000 | 0.000 | 0.00 | none | 0.000 | 0.00 | 0 | 0.00 | 32.87% | 55.02% | 87.65% | n/a | n/a | 0.534 |

results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types6_prepend_margin0p05_summary.json

@@ -0,0 +1,427 @@
+{
+  "run_root": "/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs",
+  "objective": "near_miss_policy_bc5_typedprop_p2",
+  "out_name": "policy_rollout_proposal_lattice_types6_prepend_margin0p05.json",
+  "num_completed": 3,
+  "baseline_h4_policy_success": 0.2967,
+  "baseline_h16_rank_checkpoint_success": 0.29739130434782607,
+  "mean_success": 0.32173913043478264,
+  "std_success": 0.030472026900756916,
+  "mean_progress": 0.5473962053420205,
+  "mean_action_mse_to_best": 0.4669503042795459,
+  "gain_vs_h4": 0.02503913043478262,
+  "gain_vs_h16_rank_checkpoint": 0.02434782608695657,
+  "rows": [
+    {
+      "seed": 0,
+      "path": "/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs/near_miss_policy_bc5_typedprop_p2/seed_0/policy_rollout_proposal_lattice_types6_prepend_margin0p05.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 7,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "selection_margin": 0.05,
+      "prepend_policy_candidate": true,
+      "field_optim_steps": 0,
+      "field_optim_step_size": 0.0,
+      "field_optim_trust_radius": 0.0,
+      "field_optim_l2_penalty": 0.0,
+      "retrieval_neighbors": 0,
+      "retrieval_metric": "none",
+      "retrieval_type_min_success": 0.0,
+      "retrieval_type_success_bonus_scale": 0.0,
+      "retrieval_residual_consensus_penalty_scale": 0.0,
+      "retrieval_residual_min_source_progress": 0.0,
+      "retrieval_residual_min_source_advantage": -1000000000.0,
+      "retrieval_residual_source_progress_bonus_scale": 0.0,
+      "retrieval_residual_source_score_bonus_scale": 0.0,
+      "retrieval_residual_source_advantage_bonus_scale": 0.0,
+      "retrieval_residual_composite_l2_penalty_scale": 0.0,
+      "retrieval_residual_scale": 0.0,
+      "retrieval_residual_scales": [],
+      "retrieval_residual_anchor": "none",
+      "retrieval_residual_direction": "none",
+      "retrieval_residual_reduce": "none",
+      "retrieval_residual_challenger_types": [],
+      "retrieval_residual_challenger_scales": [],
+      "retrieval_residual_challenger_margin": 0.0,
+      "lattice_exclude_types": [],
+      "candidate_type_bonuses": {},
+      "candidate_type_bonus_components": false,
+      "candidate_oracle_rollouts": 0,
+      "candidate_oracle_unique_tolerance": 1e-06,
+      "candidate_oracle_success_rate": null,
+      "candidate_oracle_progress": null,
+      "candidate_oracle_score": null,
+      "candidate_oracle_regret": null,
+      "candidate_oracle_score_gain_over_selected": null,
+      "candidate_oracle_unique_count": null,
+      "candidate_oracle_improvement_rate": null,
+      "candidate_oracle_selected_branch_success_rate": null,
+      "candidate_oracle_selected_branch_progress": null,
+      "candidate_oracle_type_counts": {},
+      "candidate_oracle_best_branch_rank": null,
+      "candidate_oracle_best_branch_rank_counts": {},
+      "candidate_oracle_branch_success_rates": [],
+      "candidate_oracle_branch_progress": [],
+      "candidate_oracle_branch_score_gains_over_selected": [],
+      "selected_residual_scale_counts": {},
+      "policy_rollout_success_rate": 0.2921739130434783,
+      "policy_rollout_progress": 0.5273328077728334,
+      "oracle_success_rate": 0.8573913043478261,
+      "action_mse_to_best": 0.42328667717110163,
+      "best_policy_val": {
+        "bc_loss": 0.1345199979841709,
+        "field_effect_loss": 0.009223711728635762,
+        "field_potential_loss": 0.2614195512400733,
+        "field_preference_loss": 0.5129645930396186,
+        "lattice_edges": 3833.3333333333335,
+        "progress_mae": 0.18681737035512924,
+        "proposal_loss": 0.15201550225416818,
+        "rank_acc": 0.8316339155038198,
+        "rank_loss": 0.5129594521390067,
+        "regret_mae": 0.3556148376729753,
+        "success_accuracy": 0.8784897228082021,
+        "total_loss": 1.8417018850644429
+      },
+      "per_task": {
+        "LiftPegUpright-v1": {
+          "action_mse_to_best": 0.3188413552863082,
+          "expert_success_rate": 0.8865979381443299,
+          "num_groups": 97,
+          "oracle_success_rate": 0.9278350515463918,
+          "policy_expert_regret": 1.2962467381634664,
+          "policy_oracle_regret": 1.33586119376507,
+          "policy_rollout_progress": 0.4537532394079818,
+          "policy_rollout_success_rate": 0.1134020618556701,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.36178428784800276,
+          "expert_success_rate": 0.9375,
+          "num_groups": 208,
+          "oracle_success_rate": 0.9471153846153846,
+          "policy_expert_regret": 1.1571640655971491,
+          "policy_oracle_regret": 1.1638306396511884,
+          "policy_rollout_progress": 0.5470511792944028,
+          "policy_rollout_success_rate": 0.22115384615384615,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.6666907884948584,
+          "expert_success_rate": 0.19480519480519481,
+          "num_groups": 77,
+          "oracle_success_rate": 0.36363636363636365,
+          "policy_expert_regret": 0.32193351367657835,
+          "policy_oracle_regret": 0.5876161599797862,
+          "policy_rollout_progress": 0.27170499244874174,
+          "policy_rollout_success_rate": 0.1038961038961039,
+          "restore_max_error": 2.3096799850463867e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.3968292265892663,
+          "expert_success_rate": 0.8617021276595744,
+          "num_groups": 94,
+          "oracle_success_rate": 0.9893617021276596,
+          "policy_expert_regret": 0.28523258579538224,
+          "policy_oracle_regret": 0.36410239719329995,
+          "policy_rollout_progress": 0.8273869645088276,
+          "policy_rollout_success_rate": 0.8085106382978723,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.4906460406538569,
+          "expert_success_rate": 0.6767676767676768,
+          "num_groups": 99,
+          "oracle_success_rate": 0.8585858585858586,
+          "policy_expert_regret": 0.9336381462788341,
+          "policy_oracle_regret": 1.0995857178261785,
+          "policy_rollout_progress": 0.4719191502140026,
+          "policy_rollout_success_rate": 0.2727272727272727,
+          "restore_max_error": 2.384185791015625e-07
+        }
+      }
+    },
+    {
+      "seed": 1,
+      "path": "/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs/near_miss_policy_bc5_typedprop_p2/seed_1/policy_rollout_proposal_lattice_types6_prepend_margin0p05.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 7,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "selection_margin": 0.05,
+      "prepend_policy_candidate": true,
+      "field_optim_steps": 0,
+      "field_optim_step_size": 0.0,
+      "field_optim_trust_radius": 0.0,
+      "field_optim_l2_penalty": 0.0,
+      "retrieval_neighbors": 0,
+      "retrieval_metric": "none",
+      "retrieval_type_min_success": 0.0,
+      "retrieval_type_success_bonus_scale": 0.0,
+      "retrieval_residual_consensus_penalty_scale": 0.0,
+      "retrieval_residual_min_source_progress": 0.0,
+      "retrieval_residual_min_source_advantage": -1000000000.0,
+      "retrieval_residual_source_progress_bonus_scale": 0.0,
+      "retrieval_residual_source_score_bonus_scale": 0.0,
+      "retrieval_residual_source_advantage_bonus_scale": 0.0,
+      "retrieval_residual_composite_l2_penalty_scale": 0.0,
+      "retrieval_residual_scale": 0.0,
+      "retrieval_residual_scales": [],
+      "retrieval_residual_anchor": "none",
+      "retrieval_residual_direction": "none",
+      "retrieval_residual_reduce": "none",
+      "retrieval_residual_challenger_types": [],
+      "retrieval_residual_challenger_scales": [],
+      "retrieval_residual_challenger_margin": 0.0,
+      "lattice_exclude_types": [],
+      "candidate_type_bonuses": {},
+      "candidate_type_bonus_components": false,
+      "candidate_oracle_rollouts": 0,
+      "candidate_oracle_unique_tolerance": 1e-06,
+      "candidate_oracle_success_rate": null,
+      "candidate_oracle_progress": null,
+      "candidate_oracle_score": null,
+      "candidate_oracle_regret": null,
+      "candidate_oracle_score_gain_over_selected": null,
+      "candidate_oracle_unique_count": null,
+      "candidate_oracle_improvement_rate": null,
+      "candidate_oracle_selected_branch_success_rate": null,
+      "candidate_oracle_selected_branch_progress": null,
+      "candidate_oracle_type_counts": {},
+      "candidate_oracle_best_branch_rank": null,
+      "candidate_oracle_best_branch_rank_counts": {},
+      "candidate_oracle_branch_success_rates": [],
+      "candidate_oracle_branch_progress": [],
+      "candidate_oracle_branch_score_gains_over_selected": [],
+      "selected_residual_scale_counts": {},
+      "policy_rollout_success_rate": 0.35304347826086957,
+      "policy_rollout_progress": 0.5706641887553016,
+      "oracle_success_rate": 0.8695652173913043,
+      "action_mse_to_best": 0.4667227768682389,
+      "best_policy_val": {
+        "bc_loss": 0.12785478805502257,
+        "field_effect_loss": 0.010064840265032318,
+        "field_potential_loss": 0.29153423342439866,
+        "field_preference_loss": 0.5177145335409377,
+        "lattice_edges": 3833.3333333333335,
+        "progress_mae": 0.19837388147910437,
+        "proposal_loss": 0.14544310627712143,
+        "rank_acc": 0.8285107778178321,
+        "rank_loss": 0.5177104257875018,
+        "regret_mae": 0.37083542181385887,
+        "success_accuracy": 0.8705372114976248,
+        "total_loss": 1.8337699174880981
+      },
+      "per_task": {
+        "LiftPegUpright-v1": {
+          "action_mse_to_best": 0.32980965103014515,
+          "expert_success_rate": 0.8584070796460177,
+          "num_groups": 113,
+          "oracle_success_rate": 0.9380530973451328,
+          "policy_expert_regret": 0.9662916387340664,
+          "policy_oracle_regret": 1.0699706825271118,
+          "policy_rollout_progress": 0.6209089936672059,
+          "policy_rollout_success_rate": 0.23008849557522124,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.4697703696725986,
+          "expert_success_rate": 0.9402173913043478,
+          "num_groups": 184,
+          "oracle_success_rate": 0.9456521739130435,
+          "policy_expert_regret": 1.0170725961508413,
+          "policy_oracle_regret": 1.0170966006257716,
+          "policy_rollout_progress": 0.6020019247065964,
+          "policy_rollout_success_rate": 0.30434782608695654,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.5883734407467082,
+          "expert_success_rate": 0.25,
+          "num_groups": 76,
+          "oracle_success_rate": 0.40789473684210525,
+          "policy_expert_regret": 0.2865761053307276,
+          "policy_oracle_regret": 0.4950702641355364,
+          "policy_rollout_progress": 0.3093867815638843,
+          "policy_rollout_success_rate": 0.17105263157894737,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.45617846688171765,
+          "expert_success_rate": 0.8198198198198198,
+          "num_groups": 111,
+          "oracle_success_rate": 1.0,
+          "policy_expert_regret": 0.4164123132422164,
+          "policy_oracle_regret": 0.4659704923092782,
+          "policy_rollout_progress": 0.7772727509339651,
+          "policy_rollout_success_rate": 0.7567567567567568,
+          "restore_max_error": 2.3795291781425476e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.5418369977192564,
+          "expert_success_rate": 0.6923076923076923,
+          "num_groups": 91,
+          "oracle_success_rate": 0.8571428571428571,
+          "policy_expert_regret": 1.0226701992553668,
+          "policy_oracle_regret": 1.1444938965551146,
+          "policy_rollout_progress": 0.4111007401248911,
+          "policy_rollout_success_rate": 0.26373626373626374,
+          "restore_max_error": 1.9371509552001953e-07
+        }
+      }
+    },
+    {
+      "seed": 2,
+      "path": "/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs/near_miss_policy_bc5_typedprop_p2/seed_2/policy_rollout_proposal_lattice_types6_prepend_margin0p05.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 7,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "selection_margin": 0.05,
+      "prepend_policy_candidate": true,
+      "field_optim_steps": 0,
+      "field_optim_step_size": 0.0,
+      "field_optim_trust_radius": 0.0,
+      "field_optim_l2_penalty": 0.0,
+      "retrieval_neighbors": 0,
+      "retrieval_metric": "none",
+      "retrieval_type_min_success": 0.0,
+      "retrieval_type_success_bonus_scale": 0.0,
+      "retrieval_residual_consensus_penalty_scale": 0.0,
+      "retrieval_residual_min_source_progress": 0.0,
+      "retrieval_residual_min_source_advantage": -1000000000.0,
+      "retrieval_residual_source_progress_bonus_scale": 0.0,
+      "retrieval_residual_source_score_bonus_scale": 0.0,
+      "retrieval_residual_source_advantage_bonus_scale": 0.0,
+      "retrieval_residual_composite_l2_penalty_scale": 0.0,
+      "retrieval_residual_scale": 0.0,
+      "retrieval_residual_scales": [],
+      "retrieval_residual_anchor": "none",
+      "retrieval_residual_direction": "none",
+      "retrieval_residual_reduce": "none",
+      "retrieval_residual_challenger_types": [],
+      "retrieval_residual_challenger_scales": [],
+      "retrieval_residual_challenger_margin": 0.0,
+      "lattice_exclude_types": [],
+      "candidate_type_bonuses": {},
+      "candidate_type_bonus_components": false,
+      "candidate_oracle_rollouts": 0,
+      "candidate_oracle_unique_tolerance": 1e-06,
+      "candidate_oracle_success_rate": null,
+      "candidate_oracle_progress": null,
+      "candidate_oracle_score": null,
+      "candidate_oracle_regret": null,
+      "candidate_oracle_score_gain_over_selected": null,
+      "candidate_oracle_unique_count": null,
+      "candidate_oracle_improvement_rate": null,
+      "candidate_oracle_selected_branch_success_rate": null,
+      "candidate_oracle_selected_branch_progress": null,
+      "candidate_oracle_type_counts": {},
+      "candidate_oracle_best_branch_rank": null,
+      "candidate_oracle_best_branch_rank_counts": {},
+      "candidate_oracle_branch_success_rates": [],
+      "candidate_oracle_branch_progress": [],
+      "candidate_oracle_branch_score_gains_over_selected": [],
+      "selected_residual_scale_counts": {},
+      "policy_rollout_success_rate": 0.32,
+      "policy_rollout_progress": 0.5441916194979264,
+      "oracle_success_rate": 0.8765217391304347,
+      "action_mse_to_best": 0.5108414587992972,
+      "best_policy_val": {
+        "bc_loss": 0.11429175941480531,
+        "field_effect_loss": 0.009631710747877756,
+        "field_potential_loss": 0.2724720388650894,
+        "field_preference_loss": 0.5087620665629705,
+        "lattice_edges": 3833.3333333333335,
+        "progress_mae": 0.19653971824381086,
+        "proposal_loss": 0.13279944616887304,
+        "rank_acc": 0.8339865836832259,
+        "rank_loss": 0.508702975180414,
+        "regret_mae": 0.3901382436354955,
+        "success_accuracy": 0.870691223276986,
+        "total_loss": 1.710942678981357
+      },
+      "per_task": {
+        "LiftPegUpright-v1": {
+          "action_mse_to_best": 0.35052205612979986,
+          "expert_success_rate": 0.8229166666666666,
+          "num_groups": 96,
+          "oracle_success_rate": 0.9270833333333334,
+          "policy_expert_regret": 0.9310697042383254,
+          "policy_oracle_regret": 1.0250059358465176,
+          "policy_rollout_progress": 0.6349228063287834,
+          "policy_rollout_success_rate": 0.2604166666666667,
+          "restore_max_error": 1.955777406692505e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.5290472695108319,
+          "expert_success_rate": 0.9444444444444444,
+          "num_groups": 198,
+          "oracle_success_rate": 0.9595959595959596,
+          "policy_expert_regret": 0.9920152253047987,
+          "policy_oracle_regret": 0.9990464125694049,
+          "policy_rollout_progress": 0.6201360545901939,
+          "policy_rollout_success_rate": 0.32323232323232326,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.7337290021940135,
+          "expert_success_rate": 0.24444444444444444,
+          "num_groups": 90,
+          "oracle_success_rate": 0.4666666666666667,
+          "policy_expert_regret": 0.3079323196369741,
+          "policy_oracle_regret": 0.6160727765618099,
+          "policy_rollout_progress": 0.3016715893521905,
+          "policy_rollout_success_rate": 0.15555555555555556,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.38539690015339617,
+          "expert_success_rate": 0.8514851485148515,
+          "num_groups": 101,
+          "oracle_success_rate": 1.0,
+          "policy_expert_regret": 0.5709437925626736,
+          "policy_oracle_regret": 0.6459072826817485,
+          "policy_rollout_progress": 0.6907263806845882,
+          "policy_rollout_success_rate": 0.6633663366336634,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.5596851660559575,
+          "expert_success_rate": 0.7666666666666667,
+          "num_groups": 90,
+          "oracle_success_rate": 0.9111111111111111,
+          "policy_expert_regret": 1.2317535453372532,
+          "policy_oracle_regret": 1.3682888499564594,
+          "policy_rollout_progress": 0.3584093944893943,
+          "policy_rollout_success_rate": 0.15555555555555556,
+          "restore_max_error": 2.2351741790771484e-07
+        }
+      }
+    }
+  ]
+}

results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types6_prepend_margin0p05_summary.md

@@ -0,0 +1,19 @@
+# h=16 Best-Policy Checkpoint Rollout
+
+Run root: `/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs`
+Objective: `near_miss_policy_bc5_typedprop_p2`
+Result file: `policy_rollout_proposal_lattice_types6_prepend_margin0p05.json`
+Completed seeds: 3
+Baseline h=4 policy success: 29.67%
+Baseline h=16 rank-checkpoint success: 29.74%
+
+Mean success: 32.17% +/- 3.05%
+Gain vs h=16 rank checkpoint: +2.43%
+Mean progress: 54.74%
+Mean action MSE to best: 0.467
+
+| seed | mode | k | policy cand | retrieval K | retrieval metric | residual anchor | residual direction | residual reduce | min type success | type success bonus | consensus penalty | min source progress | source progress bonus | source score bonus | residual scale | residual scales | margin | sigma | opt steps | trust | success | progress | oracle | candidate oracle | oracle gain | action MSE |
+|---:|---|---:|---|---:|---|---|---|---|---:|---:|---:|---:|---:|---:|---:|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
+| 0 | proposal_lattice | 7 | yes | 0 | none | none | none | none | 0.00 | 0.000 | 0.000 | 0.00 | 0.000 | 0.000 | 0.00 | none | 0.050 | 0.00 | 0 | 0.00 | 29.22% | 52.73% | 85.74% | n/a | n/a | 0.423 |
+| 1 | proposal_lattice | 7 | yes | 0 | none | none | none | none | 0.00 | 0.000 | 0.000 | 0.00 | 0.000 | 0.000 | 0.00 | none | 0.050 | 0.00 | 0 | 0.00 | 35.30% | 57.07% | 86.96% | n/a | n/a | 0.467 |
+| 2 | proposal_lattice | 7 | yes | 0 | none | none | none | none | 0.00 | 0.000 | 0.000 | 0.00 | 0.000 | 0.000 | 0.00 | none | 0.050 | 0.00 | 0 | 0.00 | 32.00% | 54.42% | 87.65% | n/a | n/a | 0.511 |

results/paper_analysis.json

@@ -1,6 +1,6 @@
 {
   "best_clean_key": "residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmchallenger001",
-  "generated_utc": "2026-06-30T13:38:53+00:00",
+  "generated_utc": "2026-06-30T14:00:47+00:00",
   "mechanism_gap": {
     "best_clean_vs_direct_same_ckpt": 0.0776811594202898,
     "best_clean_vs_h16": 0.06318840579710144,
@@ -8131,14 +8131,202 @@
       "std_success": 0.04906690775535958
     },
     "typed_proposal_lattice_types6_prepend_margin000": {
+      "ci95_success": 0.0947578966801122,
       "label": "Typed proposal lattice head, six families, policy-prepended margin 0.00",
-      "missing": true,
-      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types6_prepend_margin0p00_summary.json"
+      "mean_action_mse_to_best": 0.4803936158428204,
+      "mean_progress": 0.5402942887342373,
+      "mean_success": 0.3130434782608696,
+      "num_completed": 3,
+      "per_task_success": {
+        "LiftPegUpright-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.19943309106022364,
+          "std_success": 0.09435399455309934
+        },
+        "PickCube-v1": {
+          "mean_num_groups": 196.66666666666666,
+          "mean_success": 0.2894834633965069,
+          "std_success": 0.052873062539821346
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.1397977266398319,
+          "std_success": 0.03381854213846167
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.7065862998030683,
+          "std_success": 0.04089883799912145
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.2111037111037111,
+          "std_success": 0.056526100397487705
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.4307826819503174,
+        "1": 0.47603451989216566,
+        "2": 0.5343636456859784
+      },
+      "seed_progress": {
+        "0": 0.5106911778385225,
+        "1": 0.5600207342295502,
+        "2": 0.5501709541346392
+      },
+      "seed_success": {
+        "0": 0.26956521739130435,
+        "1": 0.3408695652173913,
+        "2": 0.32869565217391306
+      },
+      "selected_candidate_type_counts": {
+        "policy_continuous": 332,
+        "proposal_expert": 323,
+        "proposal_near_miss": 375,
+        "proposal_no_op": 77,
+        "proposal_random_negative": 220,
+        "proposal_wrong_direction": 339,
+        "proposal_wrong_gripper": 59
+      },
+      "selected_residual_scale_counts": {},
+      "selected_type_outcomes": {
+        "policy_continuous": {
+          "count": 332.0,
+          "mean_progress": 0.5657405396530697,
+          "success_rate": 0.3493975903614458
+        },
+        "proposal_expert": {
+          "count": 323.0,
+          "mean_progress": 0.5189429238730906,
+          "success_rate": 0.23219814241486067
+        },
+        "proposal_near_miss": {
+          "count": 375.0,
+          "mean_progress": 0.5442411322593689,
+          "success_rate": 0.2826666666666667
+        },
+        "proposal_no_op": {
+          "count": 77.0,
+          "mean_progress": 0.7068199223125136,
+          "success_rate": 0.5584415584415584
+        },
+        "proposal_random_negative": {
+          "count": 220.0,
+          "mean_progress": 0.4917242307215929,
+          "success_rate": 0.3181818181818182
+        },
+        "proposal_wrong_direction": {
+          "count": 339.0,
+          "mean_progress": 0.5122972353523636,
+          "success_rate": 0.2949852507374631
+        },
+        "proposal_wrong_gripper": {
+          "count": 59.0,
+          "mean_progress": 0.6135520734271761,
+          "success_rate": 0.5084745762711864
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types6_prepend_margin0p00_summary.json",
+      "std_success": 0.0381421081729762
     },
     "typed_proposal_lattice_types6_prepend_margin005": {
+      "ci95_success": 0.07570282071459547,
       "label": "Typed proposal lattice head, six families, policy-prepended margin 0.05",
-      "missing": true,
-      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types6_prepend_margin0p05_summary.json"
+      "mean_action_mse_to_best": 0.4669503042795459,
+      "mean_progress": 0.5473962053420204,
+      "mean_success": 0.3217391304347826,
+      "num_completed": 3,
+      "per_task_success": {
+        "LiftPegUpright-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.20130240803251934,
+          "std_success": 0.07761960212436964
+        },
+        "PickCube-v1": {
+          "mean_num_groups": 196.66666666666666,
+          "mean_success": 0.2829113318243753,
+          "std_success": 0.054310646611576564
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.1435014303435356,
+          "std_success": 0.03516356428996386
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.7428779105627642,
+          "std_success": 0.07356074941947707
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.23067303067303066,
+          "std_success": 0.06520878669080932
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.42328667717110163,
+        "1": 0.4667227768682389,
+        "2": 0.5108414587992972
+      },
+      "seed_progress": {
+        "0": 0.5273328077728334,
+        "1": 0.5706641887553016,
+        "2": 0.5441916194979264
+      },
+      "seed_success": {
+        "0": 0.2921739130434783,
+        "1": 0.35304347826086957,
+        "2": 0.32
+      },
+      "selected_candidate_type_counts": {
+        "policy_continuous": 816,
+        "proposal_expert": 187,
+        "proposal_near_miss": 217,
+        "proposal_no_op": 71,
+        "proposal_random_negative": 192,
+        "proposal_wrong_direction": 188,
+        "proposal_wrong_gripper": 54
+      },
+      "selected_residual_scale_counts": {},
+      "selected_type_outcomes": {
+        "policy_continuous": {
+          "count": 816.0,
+          "mean_progress": 0.5904659853015747,
+          "success_rate": 0.35784313725490197
+        },
+        "proposal_expert": {
+          "count": 187.0,
+          "mean_progress": 0.4945241949386536,
+          "success_rate": 0.1925133689839572
+        },
+        "proposal_near_miss": {
+          "count": 217.0,
+          "mean_progress": 0.5047104929483706,
+          "success_rate": 0.2534562211981567
+        },
+        "proposal_no_op": {
+          "count": 71.0,
+          "mean_progress": 0.6820441410994865,
+          "success_rate": 0.5211267605633803
+        },
+        "proposal_random_negative": {
+          "count": 192.0,
+          "mean_progress": 0.49612928278899443,
+          "success_rate": 0.3229166666666667
+        },
+        "proposal_wrong_direction": {
+          "count": 188.0,
+          "mean_progress": 0.4412869687211957,
+          "success_rate": 0.2393617021276596
+        },
+        "proposal_wrong_gripper": {
+          "count": 54.0,
+          "mean_progress": 0.625853747267414,
+          "success_rate": 0.5185185185185185
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types6_prepend_margin0p05_summary.json",
+      "std_success": 0.030472026900756916
     }
   },
   "paired_deltas": {

results/paper_analysis.md

@@ -1,6 +1,6 @@
 # Paper Analysis
 
-Generated: `2026-06-30T13:38:53+00:00`
+Generated: `2026-06-30T14:00:47+00:00`
 
 ## Main Seed Statistics
 
@@ -60,8 +60,8 @@ Generated: `2026-06-30T13:38:53+00:00`
 | residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmchallenger001_scales035040 | K4 compatible tangents, no-op bonus 0.03, near-miss challenger gate 0.01, scale-gated 0.35/0.40 | 3 | 36.00% +/- 1.31 | +/- 3.26 | 57.37% | 0.407 | +6.26 pp |
 | residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmchallenger003 | K4 compatible tangents, no-op bonus 0.03, near-miss challenger gate 0.03 | 3 | 35.94% +/- 1.28 | +/- 3.18 | 57.36% | 0.407 | +6.20 pp |
 | residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmwgchallenger001 | K4 compatible tangents, no-op bonus 0.03, near-miss/wrong-gripper challenger gate 0.01 | 3 | 35.94% +/- 1.13 | +/- 2.81 | 57.40% | 0.424 | +6.20 pp |
-| typed_proposal_lattice_types6_prepend_margin000 | Typed proposal lattice head, six families, policy-prepended margin 0.00 | 0 | missing | missing | missing | missing | missing |
-| typed_proposal_lattice_types6_prepend_margin005 | Typed proposal lattice head, six families, policy-prepended margin 0.05 | 0 | missing | missing | missing | missing | missing |
+| typed_proposal_lattice_types6_prepend_margin000 | Typed proposal lattice head, six families, policy-prepended margin 0.00 | 3 | 31.30% +/- 3.81 | +/- 9.48 | 54.03% | 0.480 | +1.57 pp |
+| typed_proposal_lattice_types6_prepend_margin005 | Typed proposal lattice head, six families, policy-prepended margin 0.05 | 3 | 32.17% +/- 3.05 | +/- 7.57 | 54.74% | 0.467 | +2.43 pp |
 | residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus001 | K4 composed compatible tangents, no-op bonus 0.03, singleton near-miss bonus 0.01 | 3 | 35.59% +/- 0.99 | +/- 2.46 | 57.10% | 0.406 | +5.86 pp |
 | residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus002 | K4 composed compatible tangents, no-op bonus 0.03, singleton near-miss bonus 0.02 | 3 | 35.59% +/- 0.99 | +/- 2.46 | 57.10% | 0.406 | +5.86 pp |
 | residual_k4_composemasked_dropnmnoop_grid035040045_srcscorebonus002 | K4 composed type-consensus tangents, masked, drop near-miss+no-op composite, source-score bonus 0.02 | 3 | 35.48% +/- 1.22 | +/- 3.02 | 57.02% | 0.408 | +5.74 pp |

results/paper_table_status.json

@@ -1553,14 +1553,14 @@
       "story_role": "model-generated counterfactual proposal support test; field scores typed proposals rather than retrieved train-state residuals",
       "fallback_success": null,
       "pending_job": "14962264/14962356/14962363",
-      "path_exists": false,
-      "status": "pending",
-      "success": null,
-      "std_success": null,
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.3130434782608696,
+      "std_success": 0.0381421081729762,
       "completed_seeds": null,
-      "num_completed": null,
+      "num_completed": 3,
       "best_config": null,
-      "gain_vs_h16_policy": null
+      "gain_vs_h16_policy": 0.01565217391304352
     },
     {
       "key": "typed_proposal_lattice_types6_prepend_margin005",
@@ -1572,14 +1572,14 @@
       "story_role": "model-generated counterfactual proposal support test with a conservative policy-abstention margin",
       "fallback_success": null,
       "pending_job": "14962264/14962357/14962364",
-      "path_exists": false,
-      "status": "pending",
-      "success": null,
-      "std_success": null,
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.32173913043478264,
+      "std_success": 0.030472026900756916,
       "completed_seeds": null,
-      "num_completed": null,
+      "num_completed": 3,
       "best_config": null,
-      "gain_vs_h16_policy": null
+      "gain_vs_h16_policy": 0.02434782608695657
     },
     {
       "key": "retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus001",

results/paper_table_status.md

@@ -84,8 +84,8 @@ Baseline h=16 policy: 29.74%
 | retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmchallenger001_scales035040 | K4 compatible residual retrieval, near-miss challenger gate 0.01, scale-gated 0.35/0.40 | complete | 36.00% | +6.26 pp | yes | no | no | trace-motivated near-miss challenger with tangent-length reliability gating over the two shortest scales |
 | retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmchallenger003 | K4 compatible residual retrieval, near-miss challenger gate 0.03 | complete | 35.94% | +6.20 pp | yes | no | no | upper-margin sensitivity for trace-motivated singleton near-miss challenger calibration |
 | retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmwgchallenger001 | K4 compatible residual retrieval, near-miss/wrong-gripper challenger gate 0.01 | complete | 35.94% | +6.20 pp | yes | no | no | trace-motivated challenger calibration test for whether wrong-gripper residuals carry conditional selector headroom |
-| typed_proposal_lattice_types6_prepend_margin000 | Typed proposal lattice head, six families, policy-prepended margin 0.00 | pending 14962264/14962356/14962363 | pending | pending | yes | no | no | model-generated counterfactual proposal support test; field scores typed proposals rather than retrieved train-state residuals |
-| typed_proposal_lattice_types6_prepend_margin005 | Typed proposal lattice head, six families, policy-prepended margin 0.05 | pending 14962264/14962357/14962364 | pending | pending | yes | no | no | model-generated counterfactual proposal support test with a conservative policy-abstention margin |
+| typed_proposal_lattice_types6_prepend_margin000 | Typed proposal lattice head, six families, policy-prepended margin 0.00 | complete | 31.30% | +1.57 pp | yes | no | no | model-generated counterfactual proposal support test; field scores typed proposals rather than retrieved train-state residuals |
+| typed_proposal_lattice_types6_prepend_margin005 | Typed proposal lattice head, six families, policy-prepended margin 0.05 | complete | 32.17% | +2.43 pp | yes | no | no | model-generated counterfactual proposal support test with a conservative policy-abstention margin |
 | retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus001 | K4 composed compatible residual retrieval, no-op bonus 0.03, singleton near-miss bonus 0.01 | complete | 35.59% | +5.86 pp | yes | no | no | revive high-precision singleton near-miss tangents without boosting toxic near-miss+no-op composites |
 | retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus002 | K4 composed compatible residual retrieval, no-op bonus 0.03, singleton near-miss bonus 0.02 | complete | 35.59% | +5.86 pp | yes | no | no | stronger singleton near-miss revival prior on the compatible local tangent chart |
 | retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_srcscorebonus002 | K4 composed type-consensus residual retrieval, masked, drop near-miss+no-op composite, source-score bonus 0.02 | complete | 35.48% | +5.74 pp | yes | no | no | train-measured source-score prior on the compatible local tangent chart |

scripts/eval_maniskill_policy_rollout.py

@@ -86,6 +86,12 @@ def main(argv: list[str] | None = None) -> int:
         action="store_true",
         help="Prepend the current policy action as candidate 0 for lattice/retrieval_lattice.",
     )
+    parser.add_argument(
+        "--proposal-lattice-types",
+        default="",
+        help="Optional comma-separated proposal head candidate types for "
+        "selection-mode=proposal_lattice. Empty uses every type in the checkpoint.",
+    )
     parser.add_argument(
         "--field-optim-steps",
         type=int,
@@ -298,6 +304,11 @@ def main(argv: list[str] | None = None) -> int:
         for item in args.retrieval_residual_challenger_types.split(",")
         if item.strip()
     )
+    proposal_lattice_types = tuple(
+        item.strip()
+        for item in args.proposal_lattice_types.split(",")
+        if item.strip()
+    )
     try:
         retrieval_residual_challenger_scales = tuple(
             float(item.strip())
@@ -344,6 +355,7 @@ def main(argv: list[str] | None = None) -> int:
         selection_seed=args.selection_seed,
         selection_margin=args.selection_margin,
         prepend_policy_candidate=args.prepend_policy_candidate,
+        proposal_lattice_types=proposal_lattice_types,
         field_optim_steps=args.field_optim_steps,
         field_optim_step_size=args.field_optim_step_size,
         field_optim_trust_radius=args.field_optim_trust_radius,

scripts/slurm/eval_maniskill_policy_rollout.sbatch

@@ -46,6 +46,10 @@ CANDIDATE_SIGMA="${CANDIDATE_SIGMA:-0.2}"
 SELECTION_SEED="${SELECTION_SEED:-0}"
 SELECTION_MARGIN="${SELECTION_MARGIN:-0.0}"
 PREPEND_POLICY_CANDIDATE="${PREPEND_POLICY_CANDIDATE:-0}"
+PROPOSAL_LATTICE_TYPES="${PROPOSAL_LATTICE_TYPES:-}"
+if [[ -n "${PROPOSAL_LATTICE_TYPES_COLON:-}" ]]; then
+  PROPOSAL_LATTICE_TYPES="${PROPOSAL_LATTICE_TYPES_COLON//:/,}"
+fi
 FIELD_OPTIM_STEPS="${FIELD_OPTIM_STEPS:-0}"
 FIELD_OPTIM_STEP_SIZE="${FIELD_OPTIM_STEP_SIZE:-0.05}"
 FIELD_OPTIM_TRUST_RADIUS="${FIELD_OPTIM_TRUST_RADIUS:-0.5}"
@@ -136,6 +140,7 @@ apptainer exec --nv \
     --candidate-sigma "$CANDIDATE_SIGMA" \
     --selection-seed "$SELECTION_SEED" \
     --selection-margin "$SELECTION_MARGIN" \
+    --proposal-lattice-types "$PROPOSAL_LATTICE_TYPES" \
     --field-optim-steps "$FIELD_OPTIM_STEPS" \
     --field-optim-step-size "$FIELD_OPTIM_STEP_SIZE" \
     --field-optim-trust-radius "$FIELD_OPTIM_TRUST_RADIUS" \

scripts/slurm/summarize_h16_policy_ckpt.sbatch

@@ -62,6 +62,7 @@ for result_path in sorted(base_dir.glob(f"seed_*/{out_name}")):
             "num_candidates": data.get("num_candidates"),
             "candidate_sigma": data.get("candidate_sigma"),
             "proposal_types": data.get("proposal_types", []),
+            "proposal_lattice_types": data.get("proposal_lattice_types", []),
             "selection_margin": data.get("selection_margin", 0.0),
             "prepend_policy_candidate": data.get("prepend_policy_candidate", False),
             "field_optim_steps": data.get("field_optim_steps", 0),