Auto-sync: 2026-06-30 10:19:50 (part 2)

results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_nooponly_prepend_margin0p05_summary.json

@@ -0,0 +1,436 @@
+{
+  "run_root": "/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs",
+  "objective": "near_miss_policy_bc5_typedprop_p2",
+  "out_name": "policy_rollout_proposal_lattice_nooponly_prepend_margin0p05.json",
+  "num_completed": 3,
+  "baseline_h4_policy_success": 0.2967,
+  "baseline_h16_rank_checkpoint_success": 0.29739130434782607,
+  "mean_success": 0.3443478260869565,
+  "std_success": 0.04558379972897795,
+  "mean_progress": 0.5623435350146948,
+  "mean_action_mse_to_best": 0.44973472340043424,
+  "gain_vs_h4": 0.0476478260869565,
+  "gain_vs_h16_rank_checkpoint": 0.04695652173913045,
+  "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_nooponly_prepend_margin0p05.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 2,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "proposal_lattice_types": [
+        "no_op"
+      ],
+      "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": 0.29391304347826086,
+      "policy_rollout_progress": 0.5316033441605775,
+      "oracle_success_rate": 0.8573913043478261,
+      "action_mse_to_best": 0.4108021359785419,
+      "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": 0.3201369491511399,
+          "expert_success_rate": 0.8865979381443299,
+          "num_groups": 97,
+          "oracle_success_rate": 0.9278350515463918,
+          "policy_expert_regret": 1.2831320579826218,
+          "policy_oracle_regret": 1.3227465135842253,
+          "policy_rollout_progress": 0.46686791958882634,
+          "policy_rollout_success_rate": 0.1134020618556701,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.3549191451982971,
+          "expert_success_rate": 0.9375,
+          "num_groups": 208,
+          "oracle_success_rate": 0.9471153846153846,
+          "policy_expert_regret": 1.1682198462792888,
+          "policy_oracle_regret": 1.1748864203333282,
+          "policy_rollout_progress": 0.5456107832276478,
+          "policy_rollout_success_rate": 0.21153846153846154,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.632519160528885,
+          "expert_success_rate": 0.19480519480519481,
+          "num_groups": 77,
+          "oracle_success_rate": 0.36363636363636365,
+          "policy_expert_regret": 0.3061711120044256,
+          "policy_oracle_regret": 0.5809461816951826,
+          "policy_rollout_progress": 0.27868307227051103,
+          "policy_rollout_success_rate": 0.14285714285714285,
+          "restore_max_error": 2.3096799850463867e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.3663194225110272,
+          "expert_success_rate": 0.8617021276595744,
+          "num_groups": 94,
+          "oracle_success_rate": 0.9893617021276596,
+          "policy_expert_regret": 0.24415543421785882,
+          "policy_oracle_regret": 0.3420131396740041,
+          "policy_rollout_progress": 0.8388379241557832,
+          "policy_rollout_success_rate": 0.8191489361702128,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.48683596810095503,
+          "expert_success_rate": 0.6767676767676768,
+          "num_groups": 99,
+          "oracle_success_rate": 0.8585858585858586,
+          "policy_expert_regret": 0.9334933291179965,
+          "policy_oracle_regret": 1.1110236455093732,
+          "policy_rollout_progress": 0.47059929641810333,
+          "policy_rollout_success_rate": 0.26262626262626265,
+          "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_nooponly_prepend_margin0p05.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 2,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "proposal_lattice_types": [
+        "no_op"
+      ],
+      "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": 0.3826086956521739,
+      "policy_rollout_progress": 0.5869571697699291,
+      "oracle_success_rate": 0.8695652173913043,
+      "action_mse_to_best": 0.43973633088900344,
+      "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": 0.32404863701327663,
+          "expert_success_rate": 0.8584070796460177,
+          "num_groups": 113,
+          "oracle_success_rate": 0.9380530973451328,
+          "policy_expert_regret": 0.9443869075152727,
+          "policy_oracle_regret": 1.0368707753651965,
+          "policy_rollout_progress": 0.6274015282371402,
+          "policy_rollout_success_rate": 0.25663716814159293,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.45956581966216303,
+          "expert_success_rate": 0.9402173913043478,
+          "num_groups": 184,
+          "oracle_success_rate": 0.9456521739130435,
+          "policy_expert_regret": 0.9579096352923522,
+          "policy_oracle_regret": 0.9579336397672824,
+          "policy_rollout_progress": 0.6176866246955204,
+          "policy_rollout_success_rate": 0.34782608695652173,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.5817036384756457,
+          "expert_success_rate": 0.25,
+          "num_groups": 76,
+          "oracle_success_rate": 0.40789473684210525,
+          "policy_expert_regret": 0.26425254188085856,
+          "policy_oracle_regret": 0.45188828523417834,
+          "policy_rollout_progress": 0.3256312172957941,
+          "policy_rollout_success_rate": 0.19736842105263158,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.35281119879591066,
+          "expert_success_rate": 0.8198198198198198,
+          "num_groups": 111,
+          "oracle_success_rate": 1.0,
+          "policy_expert_regret": 0.3633511696044389,
+          "policy_oracle_regret": 0.4144443819920222,
+          "policy_rollout_progress": 0.8017718342241941,
+          "policy_rollout_success_rate": 0.7837837837837838,
+          "restore_max_error": 2.3795291781425476e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.5307611414323469,
+          "expert_success_rate": 0.6923076923076923,
+          "num_groups": 91,
+          "oracle_success_rate": 0.8571428571428571,
+          "policy_expert_regret": 1.0045060436804216,
+          "policy_oracle_regret": 1.1137812864649428,
+          "policy_rollout_progress": 0.430824339226052,
+          "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_nooponly_prepend_margin0p05.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 2,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "proposal_lattice_types": [
+        "no_op"
+      ],
+      "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": 0.3565217391304348,
+      "policy_rollout_progress": 0.5684700911135777,
+      "oracle_success_rate": 0.8765217391304347,
+      "action_mse_to_best": 0.49866570333375737,
+      "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": 0.3476963357582766,
+          "expert_success_rate": 0.8229166666666666,
+          "num_groups": 96,
+          "oracle_success_rate": 0.9270833333333334,
+          "policy_expert_regret": 0.8970564791622261,
+          "policy_oracle_regret": 0.9909927107704183,
+          "policy_rollout_progress": 0.6376860314048827,
+          "policy_rollout_success_rate": 0.2916666666666667,
+          "restore_max_error": 1.955777406692505e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.5256955052178143,
+          "expert_success_rate": 0.9444444444444444,
+          "num_groups": 198,
+          "oracle_success_rate": 0.9595959595959596,
+          "policy_expert_regret": 0.9106334175707566,
+          "policy_oracle_regret": 0.917716868099465,
+          "policy_rollout_progress": 0.6491225997475211,
+          "policy_rollout_success_rate": 0.37373737373737376,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.6888830930227414,
+          "expert_success_rate": 0.24444444444444444,
+          "num_groups": 90,
+          "oracle_success_rate": 0.4666666666666667,
+          "policy_expert_regret": 0.3147414305350847,
+          "policy_oracle_regret": 0.6253484635510378,
+          "policy_rollout_progress": 0.30779078919440506,
+          "policy_rollout_success_rate": 0.17777777777777778,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.37912334848453505,
+          "expert_success_rate": 0.8514851485148515,
+          "num_groups": 101,
+          "oracle_success_rate": 1.0,
+          "policy_expert_regret": 0.41975170196873124,
+          "policy_oracle_regret": 0.530090075318176,
+          "policy_rollout_progress": 0.7471376474541013,
+          "policy_rollout_success_rate": 0.7227722772277227,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.5441698286889328,
+          "expert_success_rate": 0.7666666666666667,
+          "num_groups": 90,
+          "oracle_success_rate": 0.9111111111111111,
+          "policy_expert_regret": 1.212724201546775,
+          "policy_oracle_regret": 1.3493194088339806,
+          "policy_rollout_progress": 0.3773788356118732,
+          "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_nooponly_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_nooponly_prepend_margin0p05.json`
+Completed seeds: 3
+Baseline h=4 policy success: 29.67%
+Baseline h=16 rank-checkpoint success: 29.74%
+
+Mean success: 34.43% +/- 4.56%
+Gain vs h=16 rank checkpoint: +4.70%
+Mean progress: 56.23%
+Mean action MSE to best: 0.450
+
+| 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 | 2 | 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 | 29.39% | 53.16% | 85.74% | n/a | n/a | 0.411 |
+| 1 | proposal_lattice | 2 | 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 | 38.26% | 58.70% | 86.96% | n/a | n/a | 0.440 |
+| 2 | proposal_lattice | 2 | 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 | 35.65% | 56.85% | 87.65% | n/a | n/a | 0.499 |

results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types2sparse_prepend_margin0p05_summary.json

@@ -0,0 +1,439 @@
+{
+  "run_root": "/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs",
+  "objective": "near_miss_policy_bc5_typedprop_p2",
+  "out_name": "policy_rollout_proposal_lattice_types2sparse_prepend_margin0p05.json",
+  "num_completed": 3,
+  "baseline_h4_policy_success": 0.2967,
+  "baseline_h16_rank_checkpoint_success": 0.29739130434782607,
+  "mean_success": 0.34260869565217394,
+  "std_success": 0.041411759651445765,
+  "mean_progress": 0.5607520951499211,
+  "mean_action_mse_to_best": 0.46144251793843144,
+  "gain_vs_h4": 0.045908695652173925,
+  "gain_vs_h16_rank_checkpoint": 0.045217391304347876,
+  "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_types2sparse_prepend_margin0p05.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 3,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "proposal_lattice_types": [
+        "no_op",
+        "wrong_gripper"
+      ],
+      "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": 0.2956521739130435,
+      "policy_rollout_progress": 0.5345279813071956,
+      "oracle_success_rate": 0.8573913043478261,
+      "action_mse_to_best": 0.420117607595318,
+      "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": 0.3201369491511399,
+          "expert_success_rate": 0.8865979381443299,
+          "num_groups": 97,
+          "oracle_success_rate": 0.9278350515463918,
+          "policy_expert_regret": 1.2831320579826218,
+          "policy_oracle_regret": 1.3227465135842253,
+          "policy_rollout_progress": 0.46686791958882634,
+          "policy_rollout_success_rate": 0.1134020618556701,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.3549191451982971,
+          "expert_success_rate": 0.9375,
+          "num_groups": 208,
+          "oracle_success_rate": 0.9471153846153846,
+          "policy_expert_regret": 1.1682198462792888,
+          "policy_oracle_regret": 1.1748864203333282,
+          "policy_rollout_progress": 0.5456107832276478,
+          "policy_rollout_success_rate": 0.21153846153846154,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.6590563035615442,
+          "expert_success_rate": 0.19480519480519481,
+          "num_groups": 77,
+          "oracle_success_rate": 0.36363636363636365,
+          "policy_expert_regret": 0.27800474664220565,
+          "policy_oracle_regret": 0.5500710486591637,
+          "policy_rollout_progress": 0.3007770769975402,
+          "policy_rollout_success_rate": 0.15584415584415584,
+          "restore_max_error": 2.3096799850463867e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.39678950104466143,
+          "expert_success_rate": 0.8617021276595744,
+          "num_groups": 94,
+          "oracle_success_rate": 0.9893617021276596,
+          "policy_expert_regret": 0.24415543421785882,
+          "policy_oracle_regret": 0.3420131396740041,
+          "policy_rollout_progress": 0.8388379241557832,
+          "policy_rollout_success_rate": 0.8191489361702128,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.4913697942016609,
+          "expert_success_rate": 0.6767676767676768,
+          "num_groups": 99,
+          "oracle_success_rate": 0.8585858585858586,
+          "policy_expert_regret": 0.9336910261349245,
+          "policy_oracle_regret": 1.111221342526301,
+          "policy_rollout_progress": 0.4704015994011754,
+          "policy_rollout_success_rate": 0.26262626262626265,
+          "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_types2sparse_prepend_margin0p05.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 3,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "proposal_lattice_types": [
+        "no_op",
+        "wrong_gripper"
+      ],
+      "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": 0.3739130434782609,
+      "policy_rollout_progress": 0.5783052108326987,
+      "oracle_success_rate": 0.8695652173913043,
+      "action_mse_to_best": 0.46388629281144267,
+      "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": 0.33305917725232803,
+          "expert_success_rate": 0.8584070796460177,
+          "num_groups": 113,
+          "oracle_success_rate": 0.9380530973451328,
+          "policy_expert_regret": 0.9702966467732876,
+          "policy_oracle_regret": 1.0740347798682948,
+          "policy_rollout_progress": 0.6079366387782899,
+          "policy_rollout_success_rate": 0.23893805309734514,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.45956581966216303,
+          "expert_success_rate": 0.9402173913043478,
+          "num_groups": 184,
+          "oracle_success_rate": 0.9456521739130435,
+          "policy_expert_regret": 0.9579096352923522,
+          "policy_oracle_regret": 0.9579336397672824,
+          "policy_rollout_progress": 0.6176866246955204,
+          "policy_rollout_success_rate": 0.34782608695652173,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.5940139575165966,
+          "expert_success_rate": 0.25,
+          "num_groups": 76,
+          "oracle_success_rate": 0.40789473684210525,
+          "policy_expert_regret": 0.26535116567423467,
+          "policy_oracle_regret": 0.45244293017803056,
+          "policy_rollout_progress": 0.32458283413985844,
+          "policy_rollout_success_rate": 0.19736842105263158,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.4603107736722843,
+          "expert_success_rate": 0.8198198198198198,
+          "num_groups": 111,
+          "oracle_success_rate": 1.0,
+          "policy_expert_regret": 0.39933278337792233,
+          "policy_oracle_regret": 0.4657566951470332,
+          "policy_rollout_progress": 0.77748654809621,
+          "policy_rollout_success_rate": 0.7567567567567568,
+          "restore_max_error": 2.3795291781425476e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.5307611414323469,
+          "expert_success_rate": 0.6923076923076923,
+          "num_groups": 91,
+          "oracle_success_rate": 0.8571428571428571,
+          "policy_expert_regret": 1.0045060436804216,
+          "policy_oracle_regret": 1.1137812864649428,
+          "policy_rollout_progress": 0.430824339226052,
+          "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_types2sparse_prepend_margin0p05.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 3,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "proposal_lattice_types": [
+        "no_op",
+        "wrong_gripper"
+      ],
+      "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": 0.3582608695652174,
+      "policy_rollout_progress": 0.569423093309869,
+      "oracle_success_rate": 0.8765217391304347,
+      "action_mse_to_best": 0.5003236534085337,
+      "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": 0.3476963357582766,
+          "expert_success_rate": 0.8229166666666666,
+          "num_groups": 96,
+          "oracle_success_rate": 0.9270833333333334,
+          "policy_expert_regret": 0.8970564791622261,
+          "policy_oracle_regret": 0.9909927107704183,
+          "policy_rollout_progress": 0.6376860314048827,
+          "policy_rollout_success_rate": 0.2916666666666667,
+          "restore_max_error": 1.955777406692505e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.5256955052178143,
+          "expert_success_rate": 0.9444444444444444,
+          "num_groups": 198,
+          "oracle_success_rate": 0.9595959595959596,
+          "policy_expert_regret": 0.9106334175707566,
+          "policy_oracle_regret": 0.917716868099465,
+          "policy_rollout_progress": 0.6491225997475211,
+          "policy_rollout_success_rate": 0.37373737373737376,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.6939325481886044,
+          "expert_success_rate": 0.24444444444444444,
+          "num_groups": 90,
+          "oracle_success_rate": 0.4666666666666667,
+          "policy_expert_regret": 0.2943045587382383,
+          "policy_oracle_regret": 0.6051562398465143,
+          "policy_rollout_progress": 0.31656140860997967,
+          "policy_rollout_success_rate": 0.18888888888888888,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.38940331981618803,
+          "expert_success_rate": 0.8514851485148515,
+          "num_groups": 101,
+          "oracle_success_rate": 1.0,
+          "policy_expert_regret": 0.43955367287196734,
+          "policy_oracle_regret": 0.5320589331412079,
+          "policy_rollout_progress": 0.7451687896310693,
+          "policy_rollout_success_rate": 0.7227722772277227,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.5381764200619525,
+          "expert_success_rate": 0.7666666666666667,
+          "num_groups": 90,
+          "oracle_success_rate": 0.9111111111111111,
+          "policy_expert_regret": 1.2131966998179753,
+          "policy_oracle_regret": 1.349791907105181,
+          "policy_rollout_progress": 0.3769063373406728,
+          "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_types2sparse_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_types2sparse_prepend_margin0p05.json`
+Completed seeds: 3
+Baseline h=4 policy success: 29.67%
+Baseline h=16 rank-checkpoint success: 29.74%
+
+Mean success: 34.26% +/- 4.14%
+Gain vs h=16 rank checkpoint: +4.52%
+Mean progress: 56.08%
+Mean action MSE to best: 0.461
+
+| 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 | 3 | 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 | 29.57% | 53.45% | 85.74% | n/a | n/a | 0.420 |
+| 1 | proposal_lattice | 3 | 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 | 37.39% | 57.83% | 86.96% | n/a | n/a | 0.464 |
+| 2 | proposal_lattice | 3 | 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 | 35.83% | 56.94% | 87.65% | n/a | n/a | 0.500 |

results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types4safe_prepend_margin0p05_summary.json

@@ -0,0 +1,445 @@
+{
+  "run_root": "/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs",
+  "objective": "near_miss_policy_bc5_typedprop_p2",
+  "out_name": "policy_rollout_proposal_lattice_types4safe_prepend_margin0p05.json",
+  "num_completed": 3,
+  "baseline_h4_policy_success": 0.2967,
+  "baseline_h16_rank_checkpoint_success": 0.29739130434782607,
+  "mean_success": 0.3344927536231884,
+  "std_success": 0.0339167593653567,
+  "mean_progress": 0.554712315702902,
+  "mean_action_mse_to_best": 0.45649058701694045,
+  "gain_vs_h4": 0.0377927536231884,
+  "gain_vs_h16_rank_checkpoint": 0.03710144927536235,
+  "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_types4safe_prepend_margin0p05.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 5,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "proposal_lattice_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper"
+      ],
+      "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": 0.3008695652173913,
+      "policy_rollout_progress": 0.5334472684011511,
+      "oracle_success_rate": 0.8573913043478261,
+      "action_mse_to_best": 0.41737145260553404,
+      "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": 0.31877720901339324,
+          "expert_success_rate": 0.8865979381443299,
+          "num_groups": 97,
+          "oracle_success_rate": 0.9278350515463918,
+          "policy_expert_regret": 1.2891634000638097,
+          "policy_oracle_regret": 1.3287778556654133,
+          "policy_rollout_progress": 0.46083657750763846,
+          "policy_rollout_success_rate": 0.1134020618556701,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.3554349960851631,
+          "expert_success_rate": 0.9375,
+          "num_groups": 208,
+          "oracle_success_rate": 0.9471153846153846,
+          "policy_expert_regret": 1.1433499642880633,
+          "policy_oracle_regret": 1.1500165383421028,
+          "policy_rollout_progress": 0.5512498959881038,
+          "policy_rollout_success_rate": 0.23076923076923078,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.6396607906669499,
+          "expert_success_rate": 0.19480519480519481,
+          "num_groups": 77,
+          "oracle_success_rate": 0.36363636363636365,
+          "policy_expert_regret": 0.27980219809846446,
+          "policy_oracle_regret": 0.5512432564872426,
+          "policy_rollout_progress": 0.2909126972126496,
+          "policy_rollout_success_rate": 0.14285714285714285,
+          "restore_max_error": 2.3096799850463867e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.39673777245619196,
+          "expert_success_rate": 0.8617021276595744,
+          "num_groups": 94,
+          "oracle_success_rate": 0.9893617021276596,
+          "policy_expert_regret": 0.2654388921057924,
+          "policy_oracle_regret": 0.3443087035037102,
+          "policy_rollout_progress": 0.8365423603260771,
+          "policy_rollout_success_rate": 0.8191489361702128,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.490802881777089,
+          "expert_success_rate": 0.6767676767676768,
+          "num_groups": 99,
+          "oracle_success_rate": 0.8585858585858586,
+          "policy_expert_regret": 0.9362029040401633,
+          "policy_oracle_regret": 1.1135675610616953,
+          "policy_rollout_progress": 0.4680383170794959,
+          "policy_rollout_success_rate": 0.26262626262626265,
+          "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_types4safe_prepend_margin0p05.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 5,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "proposal_lattice_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper"
+      ],
+      "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": 0.36869565217391304,
+      "policy_rollout_progress": 0.5789500511732742,
+      "oracle_success_rate": 0.8695652173913043,
+      "action_mse_to_best": 0.4616191213037414,
+      "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": 0.33132346942443897,
+          "expert_success_rate": 0.8584070796460177,
+          "num_groups": 113,
+          "oracle_success_rate": 0.9380530973451328,
+          "policy_expert_regret": 0.9657866493263075,
+          "policy_oracle_regret": 1.06949607830132,
+          "policy_rollout_progress": 0.6213834781562333,
+          "policy_rollout_success_rate": 0.23008849557522124,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.45994694427392224,
+          "expert_success_rate": 0.9402173913043478,
+          "num_groups": 184,
+          "oracle_success_rate": 0.9456521739130435,
+          "policy_expert_regret": 0.9635133694886779,
+          "policy_oracle_regret": 0.9635373739636081,
+          "policy_rollout_progress": 0.6175176731078903,
+          "policy_rollout_success_rate": 0.3423913043478261,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.5906675589411814,
+          "expert_success_rate": 0.25,
+          "num_groups": 76,
+          "oracle_success_rate": 0.40789473684210525,
+          "policy_expert_regret": 0.31709635051849644,
+          "policy_oracle_regret": 0.5034712330977383,
+          "policy_rollout_progress": 0.2899634581745455,
+          "policy_rollout_success_rate": 0.15789473684210525,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.45576212239695024,
+          "expert_success_rate": 0.8198198198198198,
+          "num_groups": 111,
+          "oracle_success_rate": 1.0,
+          "policy_expert_regret": 0.3646993339061737,
+          "policy_oracle_regret": 0.4142575019651705,
+          "policy_rollout_progress": 0.8019587142510457,
+          "policy_rollout_success_rate": 0.7837837837837838,
+          "restore_max_error": 2.3795291781425476e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.5261634603593048,
+          "expert_success_rate": 0.6923076923076923,
+          "num_groups": 91,
+          "oracle_success_rate": 0.8571428571428571,
+          "policy_expert_regret": 1.0161659566910712,
+          "policy_oracle_regret": 1.1379896428558853,
+          "policy_rollout_progress": 0.4176049938241204,
+          "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_types4safe_prepend_margin0p05.json",
+      "num_groups": 575,
+      "selection_mode": "proposal_lattice",
+      "num_candidates": 5,
+      "candidate_sigma": 0.0,
+      "proposal_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper",
+        "wrong_direction",
+        "random_negative"
+      ],
+      "proposal_lattice_types": [
+        "expert",
+        "near_miss",
+        "no_op",
+        "wrong_gripper"
+      ],
+      "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": 0.3339130434782609,
+      "policy_rollout_progress": 0.5517396275342806,
+      "oracle_success_rate": 0.8765217391304347,
+      "action_mse_to_best": 0.490481187141546,
+      "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": 0.3500210010727945,
+          "expert_success_rate": 0.8229166666666666,
+          "num_groups": 96,
+          "oracle_success_rate": 0.9270833333333334,
+          "policy_expert_regret": 0.925410706239442,
+          "policy_oracle_regret": 1.019346937847634,
+          "policy_rollout_progress": 0.6405818043276668,
+          "policy_rollout_success_rate": 0.2604166666666667,
+          "restore_max_error": 1.955777406692505e-07
+        },
+        "PickCube-v1": {
+          "action_mse_to_best": 0.5145297171216549,
+          "expert_success_rate": 0.9444444444444444,
+          "num_groups": 198,
+          "oracle_success_rate": 0.9595959595959596,
+          "policy_expert_regret": 0.9614817717423042,
+          "policy_oracle_regret": 0.9685129590069104,
+          "policy_rollout_progress": 0.6254169792877604,
+          "policy_rollout_success_rate": 0.3484848484848485,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PullCube-v1": {
+          "action_mse_to_best": 0.6670428006909788,
+          "expert_success_rate": 0.24444444444444444,
+          "num_groups": 90,
+          "oracle_success_rate": 0.4666666666666667,
+          "policy_expert_regret": 0.27053625151101085,
+          "policy_oracle_regret": 0.5814869685098529,
+          "policy_rollout_progress": 0.33054109524107644,
+          "policy_rollout_success_rate": 0.2,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "PushCube-v1": {
+          "action_mse_to_best": 0.38501458349499373,
+          "expert_success_rate": 0.8514851485148515,
+          "num_groups": 101,
+          "oracle_success_rate": 1.0,
+          "policy_expert_regret": 0.5896868922627798,
+          "policy_oracle_regret": 0.6646511156370144,
+          "policy_rollout_progress": 0.6818835378283321,
+          "policy_rollout_success_rate": 0.6534653465346535,
+          "restore_max_error": 2.384185791015625e-07
+        },
+        "StackCube-v1": {
+          "action_mse_to_best": 0.5291939724236727,
+          "expert_success_rate": 0.7666666666666667,
+          "num_groups": 90,
+          "oracle_success_rate": 0.9111111111111111,
+          "policy_expert_regret": 1.2201245900657443,
+          "policy_oracle_regret": 1.3566656352745161,
+          "policy_rollout_progress": 0.37003260917133757,
+          "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_types4safe_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_types4safe_prepend_margin0p05.json`
+Completed seeds: 3
+Baseline h=4 policy success: 29.67%
+Baseline h=16 rank-checkpoint success: 29.74%
+
+Mean success: 33.45% +/- 3.39%
+Gain vs h=16 rank checkpoint: +3.71%
+Mean progress: 55.47%
+Mean action MSE to best: 0.456
+
+| 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 | 5 | 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 | 30.09% | 53.34% | 85.74% | n/a | n/a | 0.417 |
+| 1 | proposal_lattice | 5 | 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 | 36.87% | 57.90% | 86.96% | n/a | n/a | 0.462 |
+| 2 | proposal_lattice | 5 | 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 | 33.39% | 55.17% | 87.65% | n/a | n/a | 0.490 |

results/paper_analysis.json

@@ -1,6 +1,6 @@
 {
   "best_clean_key": "residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmchallenger001",
-  "generated_utc": "2026-06-30T14:00:47+00:00",
+  "generated_utc": "2026-06-30T14:17:14+00:00",
   "mechanism_gap": {
     "best_clean_vs_direct_same_ckpt": 0.0776811594202898,
     "best_clean_vs_h16": 0.06318840579710144,
@@ -8130,6 +8130,247 @@
       "source": "results/h16_lattice_no_expert_policy_baseline_margin000_summary.json",
       "std_success": 0.04906690775535958
     },
+    "typed_proposal_lattice_nooponly_bestpolicy_prepend_margin005": {
+      "label": "Typed proposal lattice head, no-op family only, best-policy checkpoint, policy-prepended margin 0.05",
+      "missing": true,
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpolicy_nooponly_prepend_margin0p05_summary.json"
+    },
+    "typed_proposal_lattice_nooponly_prepend_margin005": {
+      "ci95_success": 0.11324557534724171,
+      "label": "Typed proposal lattice head, no-op family only, policy-prepended margin 0.05",
+      "mean_action_mse_to_best": 0.44973472340043424,
+      "mean_progress": 0.5623435350146947,
+      "mean_success": 0.3443478260869565,
+      "num_completed": 3,
+      "per_task_success": {
+        "LiftPegUpright-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.22056863222130993,
+          "std_success": 0.0944471905070752
+        },
+        "PickCube-v1": {
+          "mean_num_groups": 196.66666666666666,
+          "mean_success": 0.31103397407745237,
+          "std_success": 0.08713418650109293
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.17266778056251741,
+          "std_success": 0.027612568085592256
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.7752349990605731,
+          "std_success": 0.04875373203573354
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.230969030969031,
+          "std_success": 0.06558956265212597
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.4108021359785419,
+        "1": 0.43973633088900344,
+        "2": 0.49866570333375737
+      },
+      "seed_progress": {
+        "0": 0.5316033441605775,
+        "1": 0.5869571697699291,
+        "2": 0.5684700911135777
+      },
+      "seed_success": {
+        "0": 0.29391304347826086,
+        "1": 0.3826086956521739,
+        "2": 0.3565217391304348
+      },
+      "selected_candidate_type_counts": {
+        "policy_continuous": 1458,
+        "proposal_no_op": 267
+      },
+      "selected_residual_scale_counts": {},
+      "selected_type_outcomes": {
+        "policy_continuous": {
+          "count": 1458.0,
+          "mean_progress": 0.5620794926518476,
+          "success_rate": 0.33127572016460904
+        },
+        "proposal_no_op": {
+          "count": 267.0,
+          "mean_progress": 0.5637853843219272,
+          "success_rate": 0.4157303370786517
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_nooponly_prepend_margin0p05_summary.json",
+      "std_success": 0.04558379972897795
+    },
+    "typed_proposal_lattice_types2sparse_bestpolicy_prepend_margin005": {
+      "label": "Typed proposal lattice head, sparse families, best-policy checkpoint, policy-prepended margin 0.05",
+      "missing": true,
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpolicy_types2sparse_prepend_margin0p05_summary.json"
+    },
+    "typed_proposal_lattice_types2sparse_prepend_margin005": {
+      "ci95_success": 0.10288081677597381,
+      "label": "Typed proposal lattice head, sparse no-op/wrong-gripper families, policy-prepended margin 0.05",
+      "mean_action_mse_to_best": 0.46144251793843144,
+      "mean_progress": 0.5607520951499211,
+      "mean_success": 0.3426086956521739,
+      "num_completed": 3,
+      "per_task_success": {
+        "LiftPegUpright-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.21466892720656064,
+          "std_success": 0.09157679938571918
+        },
+        "PickCube-v1": {
+          "mean_num_groups": 196.66666666666666,
+          "mean_success": 0.31103397407745237,
+          "std_success": 0.08713418650109293
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.18070048859522544,
+          "std_success": 0.021939771530326106
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.7662259900515641,
+          "std_success": 0.04888113011810259
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.230969030969031,
+          "std_success": 0.06558956265212597
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.420117607595318,
+        "1": 0.46388629281144267,
+        "2": 0.5003236534085337
+      },
+      "seed_progress": {
+        "0": 0.5345279813071956,
+        "1": 0.5783052108326987,
+        "2": 0.569423093309869
+      },
+      "seed_success": {
+        "0": 0.2956521739130435,
+        "1": 0.3739130434782609,
+        "2": 0.3582608695652174
+      },
+      "selected_candidate_type_counts": {
+        "policy_continuous": 1410,
+        "proposal_no_op": 246,
+        "proposal_wrong_gripper": 69
+      },
+      "selected_residual_scale_counts": {},
+      "selected_type_outcomes": {
+        "policy_continuous": {
+          "count": 1410.0,
+          "mean_progress": 0.5568550593511767,
+          "success_rate": 0.32127659574468087
+        },
+        "proposal_no_op": {
+          "count": 246.0,
+          "mean_progress": 0.5864891227092442,
+          "success_rate": 0.43902439024390244
+        },
+        "proposal_wrong_gripper": {
+          "count": 69.0,
+          "mean_progress": 0.5486290762605874,
+          "success_rate": 0.43478260869565216
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types2sparse_prepend_margin0p05_summary.json",
+      "std_success": 0.041411759651445765
+    },
+    "typed_proposal_lattice_types4safe_prepend_margin005": {
+      "ci95_success": 0.08426070119385123,
+      "label": "Typed proposal lattice head, safe four families, policy-prepended margin 0.05",
+      "mean_action_mse_to_best": 0.45649058701694045,
+      "mean_progress": 0.5547123157029019,
+      "mean_success": 0.33449275362318837,
+      "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": 0.3072151278673018,
+          "std_success": 0.06627415942419872
+        },
+        "PullCube-v1": {
+          "mean_num_groups": 81.0,
+          "mean_success": 0.1669172932330827,
+          "std_success": 0.029620624967807846
+        },
+        "PushCube-v1": {
+          "mean_num_groups": 102.0,
+          "mean_success": 0.75213268882955,
+          "std_success": 0.08725884956543115
+        },
+        "StackCube-v1": {
+          "mean_num_groups": 93.33333333333333,
+          "mean_success": 0.22730602730602734,
+          "std_success": 0.06214020979146663
+        }
+      },
+      "seed_action_mse_to_best": {
+        "0": 0.41737145260553404,
+        "1": 0.4616191213037414,
+        "2": 0.490481187141546
+      },
+      "seed_progress": {
+        "0": 0.5334472684011511,
+        "1": 0.5789500511732742,
+        "2": 0.5517396275342806
+      },
+      "seed_success": {
+        "0": 0.3008695652173913,
+        "1": 0.36869565217391304,
+        "2": 0.3339130434782609
+      },
+      "selected_candidate_type_counts": {
+        "policy_continuous": 904,
+        "proposal_expert": 217,
+        "proposal_near_miss": 314,
+        "proposal_no_op": 234,
+        "proposal_wrong_gripper": 56
+      },
+      "selected_residual_scale_counts": {},
+      "selected_type_outcomes": {
+        "policy_continuous": {
+          "count": 904.0,
+          "mean_progress": 0.5831427174690579,
+          "success_rate": 0.35508849557522126
+        },
+        "proposal_expert": {
+          "count": 217.0,
+          "mean_progress": 0.4891465714991024,
+          "success_rate": 0.19815668202764977
+        },
+        "proposal_near_miss": {
+          "count": 314.0,
+          "mean_progress": 0.4776973845615129,
+          "success_rate": 0.2484076433121019
+        },
+        "proposal_no_op": {
+          "count": 234.0,
+          "mean_progress": 0.5960542046202298,
+          "success_rate": 0.45726495726495725
+        },
+        "proposal_wrong_gripper": {
+          "count": 56.0,
+          "mean_progress": 0.6089153454772064,
+          "success_rate": 0.5
+        }
+      },
+      "source": "results/h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types4safe_prepend_margin0p05_summary.json",
+      "std_success": 0.0339167593653567
+    },
     "typed_proposal_lattice_types6_prepend_margin000": {
       "ci95_success": 0.0947578966801122,
       "label": "Typed proposal lattice head, six families, policy-prepended margin 0.00",

results/paper_analysis.md

@@ -1,6 +1,6 @@
 # Paper Analysis
 
-Generated: `2026-06-30T14:00:47+00:00`
+Generated: `2026-06-30T14:17:14+00:00`
 
 ## Main Seed Statistics
 
@@ -62,6 +62,11 @@ Generated: `2026-06-30T14:00:47+00:00`
 | 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 | 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 |
+| typed_proposal_lattice_types4safe_prepend_margin005 | Typed proposal lattice head, safe four families, policy-prepended margin 0.05 | 3 | 33.45% +/- 3.39 | +/- 8.43 | 55.47% | 0.456 | +3.71 pp |
+| typed_proposal_lattice_types2sparse_prepend_margin005 | Typed proposal lattice head, sparse no-op/wrong-gripper families, policy-prepended margin 0.05 | 3 | 34.26% +/- 4.14 | +/- 10.29 | 56.08% | 0.461 | +4.52 pp |
+| typed_proposal_lattice_nooponly_prepend_margin005 | Typed proposal lattice head, no-op family only, policy-prepended margin 0.05 | 3 | 34.43% +/- 4.56 | +/- 11.32 | 56.23% | 0.450 | +4.70 pp |
+| typed_proposal_lattice_types2sparse_bestpolicy_prepend_margin005 | Typed proposal lattice head, sparse families, best-policy checkpoint, policy-prepended margin 0.05 | 0 | missing | missing | missing | missing | missing |
+| typed_proposal_lattice_nooponly_bestpolicy_prepend_margin005 | Typed proposal lattice head, no-op family only, best-policy checkpoint, policy-prepended margin 0.05 | 0 | missing | missing | missing | missing | missing |
 | 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_core_results.md

@@ -61,7 +61,7 @@ lattice is `+27.25 pp`, and the remaining clean-to-same-state proposal gap is
 | K4 compatible residual retrieval, unique candidate-oracle prefix K=8 | No | No | 43.07% diagnostic | +13.33 pp diagnostic | Diagnostic-only measured oracle over generated clean candidate prefix; unique-action trace shows real selector headroom, not duplicate-action inflation |
 | K4 compatible residual retrieval + near-miss challenger gate 0.01 | No | No | 36.06% | +6.32 pp | Current best clean deployment row; a two-stage selector keeps the robust compatible-chart anchor and only lets singleton near-miss tangents override under a tightly calibrated positive field margin |
 | K4 compatible residual retrieval + near-miss challenger fine calibration | No | No | 36.00-36.00% | +6.26 pp | Margins 0.005/0.015/0.02 form a near-tie plateau around the 0.01 optimum; scale-gating the challenger to 0.35 or 0.35/0.40 lowers MSE slightly but does not recover the lost success; 0.03 and near-miss+wrong-gripper are lower or higher-MSE, so the effect is calibrated singleton near-miss geometry rather than broad residual mixing |
-| Typed proposal lattice head, six primitive families | No | No | pending | pending | New clean support-gap test: the model generates expert/non-expert primitive proposals directly and the field scores them with a policy fallback; train/eval/summary chain is `14962264` -> `14962356`/`14962357` -> `14962363`/`14962364` |
+| Typed proposal lattice head, generated primitive families | No | No | 31.30-34.43% | +1.57-4.70 pp | Direct generated support improves as anti-goal families are masked: six-family is 31.30%/32.17%, safe4 is 33.45%, sparse no-op/wrong-gripper is 34.26%, and no-op-only is 34.43%; still below compatible tangent transport, so proposal locality remains the bottleneck. Best-policy checkpoint ablations are pending as `14963143`/`14963147` |
 | K4 compatible residual retrieval, margin 0.10 | No | No | 34.67% | +4.93 pp | Naively lowering abstention selects too many bad nonzero tangents and falls below the 36.06% top row |
 | K4 composed type-consensus residual retrieval, exact compatibility mask + source-score prior | No | No | 35.48% | +5.74 pp | Measured train-source reward confidence does not replace the sparse typed prior on the compatible chart; useful negative calibration |
 | K4 composed type-consensus residual retrieval, exact compatibility mask + no-op/source-score priors | No | No | 35.54% | +5.80 pp | Adding measured train-source reward confidence to the current typed-prior chart ties near-best but does not beat exact compatibility masking alone |
@@ -116,10 +116,11 @@ Suggested main-table rows:
 15. K4 masked composed type-consensus residual retrieval + no-op prior 0.03
 16. K4 masked composed type-consensus residual retrieval + exact near-miss/no-op compatibility mask, with and without typed no-op prior
 17. K4 compatible tangent near-miss challenger gate 0.01 and fine margin calibration
-18. K4 masked composed type-consensus residual retrieval + source-score prior on the exact compatibility chart
-19. K4 masked composed type-consensus residual retrieval + component-wise no-op prior diagnostic
-20. K4 masked composed type-consensus residual retrieval + composite trust penalty diagnostic
-20. K4 mean-by-type residual retrieval + upper/wide tangent-length diagnostics
+18. Typed proposal lattice generated-family support-gap diagnostic
+19. K4 masked composed type-consensus residual retrieval + source-score prior on the exact compatibility chart
+20. K4 masked composed type-consensus residual retrieval + component-wise no-op prior diagnostic
+21. K4 masked composed type-consensus residual retrieval + composite trust penalty diagnostic
+22. K4 mean-by-type residual retrieval + upper/wide tangent-length diagnostics
 21. K4 mean-by-type residual retrieval + minimum-energy action penalty diagnostics
 22. K4 mean-by-type residual retrieval + source-progress/source-score/source-advantage prior diagnostics
 23. K4 mean-by-type residual retrieval + train-family success bonus diagnostics

results/paper_story_memo.md

@@ -34,7 +34,7 @@ when queried on proposal geometry that matches those local counterfactuals.
 | Singleton near-miss priors do not open a new clean route | adding a small exact `residual_near_miss` singleton prior 0.01 or 0.02 on the compatible chart ties the previous 35.59% compatibility row and slightly raises progress to 57.10% | Tie diagnostic: revived near-miss tangents are high-precision but too sparse to close the clean-to-same-state proposal gap |
 | Candidate-prefix oracle needs unique-action hygiene | the first K=8 prefix diagnostic was archived as `_nonunique`; the deduplicated unique-action trace reaches 43.07% candidate-oracle success with mean best branch rank 2.85 | Supported diagnostic: proposal headroom is real, but branch success falls with rank, so selector calibration must be conditional |
 | Near-miss challenger calibration improves clean deployment | the trace-motivated two-stage selector keeps the compatible-chart anchor and lets singleton near-miss tangents override only under a tightly calibrated margin; margin 0.01 reaches 36.06%, while 0.005/0.015/0.02 and scale-gated 0.35 or 0.35/0.40 variants tie at 36.00% | Current best clean row; small but story-aligned selector-calibration gain, with scale-gating explaining MSE but not improving success and wrong-gripper challengers rejected by lower success or higher MSE |
-| Typed proposal generation tests the remaining support gap | a new optional proposal head predicts expert plus primitive non-expert action families, then `proposal_lattice` lets the field score the generated set with policy fallback; jobs `14962264` -> `14962356`/`14962357` -> `14962363`/`14962364` are pending | Hypothesis test: if clean proposal support is the bottleneck, a learned typed lattice should close part of the 20.93 pp clean-to-same-state gap without same-state candidate leakage |
+| Typed proposal generation confirms support geometry, not just support count, is the remaining clean bottleneck | six-family generated typed proposal lattices reach 31.30%/32.17%; masking anti-goal families raises safe4 to 33.45%, sparse no-op/wrong-gripper to 34.26%, and no-op-only to 34.43%, still below the 36.06% compatible-tangent row | Negative/diagnostic: direct proposal generation helps only when sparse and anti-goal masked, but learned proposals still lack the local same-state counterfactual geometry |
 | Component-wise composite priors do not add the gain | propagating the no-op bonus into composite types reaches 35.36%, below the exact-prior masked composition row at 35.54% and the exact compatibility row at 35.59% | Negative/near-tie diagnostic: sparse exact priors are better than broadly rewarding every no-op-containing composite |
 | Composite trust-radius penalty explains but does not improve the success top line | composite-only L2 penalty 0.02 ties 35.54% while lowering action MSE from 0.4106 to 0.4079; with the exact compatibility mask it lowers MSE further to 0.4048 but drops back to 35.54% | Tie/negative diagnostic: composed tangents need a local trust radius, but compatibility masking gives the success optimum |
 | Minimum-energy residual regularization does not add the gain | action L2 penalty 0.05 ties the previous 35.42% scale-grid row, while 0.10/0.20 reach 35.36% | Negative/tie diagnostic: the clean bridge is not explained by shortest-action bias |
@@ -85,8 +85,9 @@ clean proposal result, the intended main rows are:
 17. K4 masked composed type-consensus tangent transport: 35.30%; with exact near-miss/no-op compatibility mask only: 35.48%; with typed no-op prior: 35.54%; with exact compatibility mask + typed no-op prior: 35.59%; adding singleton near-miss prior 0.01/0.02 ties 35.59%; with source-score prior on that exact chart: 35.48%; with typed no-op + source-score: 35.54%
 18. K4 compatible tangent unique-action candidate-oracle prefix K=8 diagnostic: 43.07%; branch trace shows mean best branch rank 2.85, unique count 8.00, and real but conditional selector headroom
 19. K4 compatible tangent near-miss challenger gate 0.01: 36.06%; current best clean deployment row from trace-motivated two-stage selector calibration. Fine margins 0.005/0.015/0.02 tie at 36.00%; scale-gating the challenger to 0.35 or 0.35/0.40 also ties 36.00% with slightly lower MSE; 0.03 and near-miss+wrong-gripper variants are lower or higher-MSE.
-20. K4 compatible tangent abstention margin 0.10 diagnostic: 34.67%; naive lower abstention hurts despite candidate-prefix oracle headroom
-20. K4 masked composed type-consensus tangent transport + component-wise no-op prior: 35.36%
+20. Typed proposal lattice head: six-family generated proposals reach 31.30%/32.17%; safe4 reaches 33.45%; sparse no-op/wrong-gripper reaches 34.26%; no-op-only reaches 34.43%. Checkpointing ablations with `best_policy.pt` are pending as `14963143`/`14963147`.
+21. K4 compatible tangent abstention margin 0.10 diagnostic: 34.67%; naive lower abstention hurts despite candidate-prefix oracle headroom
+22. K4 masked composed type-consensus tangent transport + component-wise no-op prior: 35.36%
 21. K4 masked composed type-consensus tangent transport + composite L2 trust penalty: 35.54% at 0.02; 35.48% at 0.05; 35.54% with exact compatibility mask + L2 0.02
 22. K4 mean-by-type tangent consensus + upper/wide scale diagnostics: 35.36% for 0.40/0.45/0.50; 35.13% for 0.35/0.45/0.55
 23. K4 mean-by-type tangent consensus + action L2 penalty: 35.42% at 0.05; 35.36% at 0.10/0.20

results/paper_table_status.json

@@ -1581,6 +1581,101 @@
       "best_config": null,
       "gain_vs_h16_policy": 0.02434782608695657
     },
+    {
+      "key": "typed_proposal_lattice_types4safe_prepend_margin005",
+      "label": "Typed proposal lattice head, safe four families, policy-prepended margin 0.05",
+      "path": "h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types4safe_prepend_margin0p05_summary.json",
+      "clean_deployment": "yes",
+      "same_state_proposals": "no",
+      "expert_proposal": "no",
+      "story_role": "anti-goal proposal-family mask test for the model-generated typed lattice",
+      "fallback_success": null,
+      "pending_job": "14963030/14963044",
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.3344927536231884,
+      "std_success": 0.0339167593653567,
+      "completed_seeds": null,
+      "num_completed": 3,
+      "best_config": null,
+      "gain_vs_h16_policy": 0.03710144927536235
+    },
+    {
+      "key": "typed_proposal_lattice_types2sparse_prepend_margin005",
+      "label": "Typed proposal lattice head, sparse no-op/wrong-gripper families, policy-prepended margin 0.05",
+      "path": "h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_types2sparse_prepend_margin0p05_summary.json",
+      "clean_deployment": "yes",
+      "same_state_proposals": "no",
+      "expert_proposal": "no",
+      "story_role": "sparse safe-support test for whether typed generated no-op/wrong-gripper proposals recover the clean support gap",
+      "fallback_success": null,
+      "pending_job": "14963031/14963042",
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.34260869565217394,
+      "std_success": 0.041411759651445765,
+      "completed_seeds": null,
+      "num_completed": 3,
+      "best_config": null,
+      "gain_vs_h16_policy": 0.045217391304347876
+    },
+    {
+      "key": "typed_proposal_lattice_nooponly_prepend_margin005",
+      "label": "Typed proposal lattice head, no-op family only, policy-prepended margin 0.05",
+      "path": "h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpt_proposal_lattice_nooponly_prepend_margin0p05_summary.json",
+      "clean_deployment": "yes",
+      "same_state_proposals": "no",
+      "expert_proposal": "no",
+      "story_role": "minimal generated no-op proposal-family test after sparse typed lattice improved over the full six-family row",
+      "fallback_success": null,
+      "pending_job": "14963032/14963043",
+      "path_exists": true,
+      "status": "complete",
+      "success": 0.3443478260869565,
+      "std_success": 0.04558379972897795,
+      "completed_seeds": null,
+      "num_completed": 3,
+      "best_config": null,
+      "gain_vs_h16_policy": 0.04695652173913045
+    },
+    {
+      "key": "typed_proposal_lattice_types2sparse_bestpolicy_prepend_margin005",
+      "label": "Typed proposal lattice head, sparse no-op/wrong-gripper families, best-policy checkpoint",
+      "path": "h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpolicy_types2sparse_prepend_margin0p05_summary.json",
+      "clean_deployment": "yes",
+      "same_state_proposals": "no",
+      "expert_proposal": "no",
+      "story_role": "checkpointing ablation for typed generated sparse proposals: lower BC/proposal loss versus weaker field scorer",
+      "fallback_success": null,
+      "pending_job": "14963143/14963150",
+      "path_exists": false,
+      "status": "pending",
+      "success": null,
+      "std_success": null,
+      "completed_seeds": null,
+      "num_completed": null,
+      "best_config": null,
+      "gain_vs_h16_policy": null
+    },
+    {
+      "key": "typed_proposal_lattice_nooponly_bestpolicy_prepend_margin005",
+      "label": "Typed proposal lattice head, no-op family only, best-policy checkpoint",
+      "path": "h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpolicy_nooponly_prepend_margin0p05_summary.json",
+      "clean_deployment": "yes",
+      "same_state_proposals": "no",
+      "expert_proposal": "no",
+      "story_role": "minimal generated no-op checkpointing ablation for typed proposal support",
+      "fallback_success": null,
+      "pending_job": "14963147/14963149",
+      "path_exists": false,
+      "status": "pending",
+      "success": null,
+      "std_success": null,
+      "completed_seeds": null,
+      "num_completed": null,
+      "best_config": null,
+      "gain_vs_h16_policy": null
+    },
     {
       "key": "retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus001",
       "label": "K4 composed compatible residual retrieval, no-op bonus 0.03, singleton near-miss bonus 0.01",

results/paper_table_status.md

@@ -86,6 +86,11 @@ Baseline h=16 policy: 29.74%
 | 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 | 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 |
+| typed_proposal_lattice_types4safe_prepend_margin005 | Typed proposal lattice head, safe four families, policy-prepended margin 0.05 | complete | 33.45% | +3.71 pp | yes | no | no | anti-goal proposal-family mask test for the model-generated typed lattice |
+| typed_proposal_lattice_types2sparse_prepend_margin005 | Typed proposal lattice head, sparse no-op/wrong-gripper families, policy-prepended margin 0.05 | complete | 34.26% | +4.52 pp | yes | no | no | sparse safe-support test for whether typed generated no-op/wrong-gripper proposals recover the clean support gap |
+| typed_proposal_lattice_nooponly_prepend_margin005 | Typed proposal lattice head, no-op family only, policy-prepended margin 0.05 | complete | 34.43% | +4.70 pp | yes | no | no | minimal generated no-op proposal-family test after sparse typed lattice improved over the full six-family row |
+| typed_proposal_lattice_types2sparse_bestpolicy_prepend_margin005 | Typed proposal lattice head, sparse no-op/wrong-gripper families, best-policy checkpoint | pending 14963143/14963150 | pending | pending | yes | no | no | checkpointing ablation for typed generated sparse proposals: lower BC/proposal loss versus weaker field scorer |
+| typed_proposal_lattice_nooponly_bestpolicy_prepend_margin005 | Typed proposal lattice head, no-op family only, best-policy checkpoint | pending 14963147/14963149 | pending | pending | yes | no | no | minimal generated no-op checkpointing ablation for typed proposal support |
 | 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/build_paper_analysis.py

@@ -497,6 +497,22 @@ METHODS = [
             "proposal_lattice_nooponly_prepend_margin0p05_summary.json"
         ),
     ),
+    MethodSpec(
+        key="typed_proposal_lattice_types2sparse_bestpolicy_prepend_margin005",
+        label="Typed proposal lattice head, sparse families, best-policy checkpoint, policy-prepended margin 0.05",
+        summary_path=(
+            "h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_"
+            "bestpolicy_types2sparse_prepend_margin0p05_summary.json"
+        ),
+    ),
+    MethodSpec(
+        key="typed_proposal_lattice_nooponly_bestpolicy_prepend_margin005",
+        label="Typed proposal lattice head, no-op family only, best-policy checkpoint, policy-prepended margin 0.05",
+        summary_path=(
+            "h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_"
+            "bestpolicy_nooponly_prepend_margin0p05_summary.json"
+        ),
+    ),
     MethodSpec(
         key="residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus001",
         label="K4 composed compatible tangents, no-op bonus 0.03, singleton near-miss bonus 0.01",

scripts/build_paper_table_status.py

@@ -875,6 +875,26 @@ SPECS = [
         story_role="minimal generated no-op proposal-family test after sparse typed lattice improved over the full six-family row",
         pending_job="14963032/14963043",
     ),
+    ResultSpec(
+        key="typed_proposal_lattice_types2sparse_bestpolicy_prepend_margin005",
+        label="Typed proposal lattice head, sparse no-op/wrong-gripper families, best-policy checkpoint",
+        path="h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpolicy_types2sparse_prepend_margin0p05_summary.json",
+        clean_deployment="yes",
+        same_state_proposals="no",
+        expert_proposal="no",
+        story_role="checkpointing ablation for typed generated sparse proposals: lower BC/proposal loss versus weaker field scorer",
+        pending_job="14963143/14963150",
+    ),
+    ResultSpec(
+        key="typed_proposal_lattice_nooponly_bestpolicy_prepend_margin005",
+        label="Typed proposal lattice head, no-op family only, best-policy checkpoint",
+        path="h16_policy_ckpt_near_miss_policy_bc5_typedprop_p2_bestpolicy_nooponly_prepend_margin0p05_summary.json",
+        clean_deployment="yes",
+        same_state_proposals="no",
+        expert_proposal="no",
+        story_role="minimal generated no-op checkpointing ablation for typed proposal support",
+        pending_job="14963147/14963149",
+    ),
     ResultSpec(
         key="retrieval_residual_k4_composemasked_dropnmnoop_grid035040045_noopbonus003_nmbonus001",
         label="K4 composed compatible residual retrieval, no-op bonus 0.03, singleton near-miss bonus 0.01",