results: retain failed oral B confirmation gate

1 files changed, 248 insertions, 0 deletions

diff --git a/results/bci_td_confirmation/bci_td_confirm_t13_m3.json b/results/bci_td_confirmation/bci_td_confirm_t13_m3.json
new file mode 100644
index 0000000..0fc9e3f
--- /dev/null
+++ b/results/bci_td_confirmation/bci_td_confirm_t13_m3.json
@@ -0,0 +1,248 @@
+{
+  "args": {
+    "d4_gate": "results/kp_dynamic_projection_confirmation_gate.json",
+    "model_seed": 3,
+    "outdir": "results/bci_td_confirmation",
+    "r1_gate": "results/bci_td_dev_gate.json",
+    "task_seed": 13
+  },
+  "conditions": {
+    "fixed_vectorizer": {
+      "cost": {
+        "conservative_online_cursor_scalar_observations": 0,
+        "online_role_perturbation_events": 98,
+        "ordinary_performance_change_observations": 25088,
+        "ordinary_state_episode_steps": 25088
+      },
+      "daily_success": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "early_success": 0.0,
+      "final_success": 0.0,
+      "late_success": 0.0,
+      "learning_gain": 0.0,
+      "role_cosine_after_training": -0.006882439833134413,
+      "training_wall_s": 0.13782113417983055
+    },
+    "intact": {
+      "cost": {
+        "conservative_online_cursor_scalar_observations": 12544,
+        "online_role_perturbation_events": 98,
+        "ordinary_performance_change_observations": 25088,
+        "ordinary_state_episode_steps": 25088
+      },
+      "daily_success": [
+        0.015625,
+        0.078125,
+        0.328125,
+        0.671875,
+        0.75,
+        0.765625,
+        0.84375,
+        0.96875,
+        0.921875,
+        0.953125,
+        1.0,
+        1.0,
+        1.0,
+        1.0
+      ],
+      "early_success": 0.140625,
+      "final_success": 0.99609375,
+      "late_success": 1.0,
+      "learning_gain": 0.859375,
+      "role_cosine_after_training": 0.9341226816177368,
+      "training_wall_s": 0.19381169974803925
+    },
+    "oracle_role": {
+      "cost": {
+        "conservative_online_cursor_scalar_observations": 0,
+        "online_role_perturbation_events": 98,
+        "ordinary_performance_change_observations": 25088,
+        "ordinary_state_episode_steps": 25088
+      },
+      "daily_success": [
+        0.015625,
+        0.078125,
+        0.34375,
+        0.671875,
+        0.75,
+        0.765625,
+        0.84375,
+        0.96875,
+        0.953125,
+        0.953125,
+        1.0,
+        1.0,
+        1.0,
+        1.0
+      ],
+      "early_success": 0.14583333333333334,
+      "final_success": 0.9921875,
+      "late_success": 1.0,
+      "learning_gain": 0.8541666666666666,
+      "role_cosine_after_training": 1.0000001192092896,
+      "training_wall_s": 0.13274051621556282
+    },
+    "plasticity_lesion": {
+      "cost": {
+        "conservative_online_cursor_scalar_observations": 12544,
+        "online_role_perturbation_events": 98,
+        "ordinary_performance_change_observations": 25088,
+        "ordinary_state_episode_steps": 25088
+      },
+      "daily_success": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.015625,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.015625,
+        0.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "early_success": 0.0,
+      "final_success": 0.0,
+      "late_success": 0.0,
+      "learning_gain": 0.0,
+      "role_cosine_after_training": 0.9944544434547424,
+      "training_wall_s": 0.11727873980998993
+    }
+  },
+  "config": {
+    "context_ar": 0.8,
+    "context_dim": 16,
+    "coupling_scale": 1.0,
+    "days": 14,
+    "episodes_per_day": 64,
+    "feedback": "performance_velocity",
+    "forward_eta": 0.1,
+    "inertia": 0.65,
+    "kappa": 0.0,
+    "n_background": 30,
+    "n_minus": 5,
+    "n_plus": 5,
+    "perturb_every": 4,
+    "perturb_sigma": 0.03,
+    "predictor_eta": 0.2,
+    "process_noise": 0.12,
+    "steps_per_episode": 28,
+    "target": 0.8,
+    "vectorizer_eta": 0.03
+  },
+  "finite": true,
+  "hardware": {
+    "device": "cpu",
+    "platform": "Linux-5.15.0-161-generic-x86_64-with-glibc2.35",
+    "threads": 1,
+    "torch_version": "2.3.1+cu118"
+  },
+  "peak_rss_mib": 411.17578125,
+  "protocol": {
+    "confirmation_grid_size": 30,
+    "d4_gate_sha256": "636c587e47287338cdca7d9558dc3bfb99a2cec615badb23337025d85b8128ef",
+    "evaluation_task_seed": 300013,
+    "name": "oral_b_td_confirmation_v1",
+    "no_further_selection": true,
+    "protocol_sha256": "3a1514e851e25e6f2588a46638992e3cb5530eb71ad79f45bb573d3a701d98bd",
+    "r1_gate_sha256": "ebb28c57461dae8334f623743145ad0a934a8fc14a96a1d877992b74daec6c8a",
+    "selected_eta": 0.1,
+    "split": "untouched_confirmation",
+    "training_task_seed": 13
+  },
+  "provenance": {
+    "git_commit": "1ff7cb7cf7cecd79cf5a0ab36c149342482c7005",
+    "git_tracked_dirty": false,
+    "input_sha256": {
+      "d4_gate": "636c587e47287338cdca7d9558dc3bfb99a2cec615badb23337025d85b8128ef",
+      "development_runner": "71f3d15482c5f96bf1d18297f0ea29181dd6f00a527f2e2ace07dc9b77489dc2",
+      "protocol": "3a1514e851e25e6f2588a46638992e3cb5530eb71ad79f45bb573d3a701d98bd",
+      "r1_gate": "ebb28c57461dae8334f623743145ad0a934a8fc14a96a1d877992b74daec6c8a",
+      "runner": "5fd0735d9912923980b4b3b72756a630f76aceffc8c3618f1c852a7f936b2f05"
+    },
+    "tracked_inputs": {
+      "d4_gate": true,
+      "development_runner": true,
+      "protocol": true,
+      "r1_gate": true,
+      "runner": true
+    }
+  },
+  "schema_version": 1,
+  "signatures": {
+    "active_training_events": 11786,
+    "causal_role_sign_inversion_index": 0.07642599797607735,
+    "decoder_distance_residual_corr": 0.16901581489180523,
+    "early_residual_late_activity_change_corr": -0.6335484767130429,
+    "evaluation_episodes": 256,
+    "evaluation_success_fraction": 0.99609375,
+    "mean_abs_raw_soma_corr": 0.9976102168085845,
+    "mean_abs_residual_soma_corr": 0.06203723643043235,
+    "raw_minus_residual_abs_soma_corr": 0.9355729803781522,
+    "residual_minus_soma_outcome_acc": -0.04509803921568628,
+    "residual_outcome_decoder_balanced_acc": 0.4549019607843137,
+    "role_aligned_error_cv_corr": 0.3051155204031925,
+    "role_aligned_velocity_cv_corr": 0.9424342189758789,
+    "soma_outcome_decoder_balanced_acc": 0.5,
+    "surrounding_event_decoder_balanced_acc": 0.5521487384411433,
+    "velocity_minus_error_abs_cv_corr": 0.6373186985726864
+  },
+  "wall_s": 1.135385800153017,
+  "warmup": {
+    "fixed_vectorizer": {
+      "batches": 100,
+      "examples": 6400,
+      "instruction_present": false,
+      "predictor_max_abs_error": 2.384185791015625e-07,
+      "rng_seed": 401303,
+      "role_cosine_after_warmup": -0.006882439833134413,
+      "role_cursor_scalar_observations": 12800
+    },
+    "intact": {
+      "batches": 100,
+      "examples": 6400,
+      "instruction_present": false,
+      "predictor_max_abs_error": 2.384185791015625e-07,
+      "rng_seed": 401303,
+      "role_cosine_after_warmup": 0.9951484799385071,
+      "role_cursor_scalar_observations": 12800
+    },
+    "oracle_role": {
+      "batches": 100,
+      "examples": 6400,
+      "instruction_present": false,
+      "predictor_max_abs_error": 2.384185791015625e-07,
+      "rng_seed": 401303,
+      "role_cosine_after_warmup": 1.0000001192092896,
+      "role_cursor_scalar_observations": 12800
+    },
+    "plasticity_lesion": {
+      "batches": 100,
+      "examples": 6400,
+      "instruction_present": false,
+      "predictor_max_abs_error": 2.384185791015625e-07,
+      "rng_seed": 401303,
+      "role_cosine_after_warmup": 0.9951484799385071,
+      "role_cursor_scalar_observations": 12800
+    }
+  }
+}