results: close stagewise no-KP feedback screen

4 files changed, 750 insertions, 0 deletions

diff --git a/ORAL_A_V6.md b/ORAL_A_V6.md
index 7663f27..6e597f4 100644
--- a/ORAL_A_V6.md
+++ b/ORAL_A_V6.md
@@ -94,3 +94,32 @@ recovery after the endpoint.  Failure closes V6 and the present no-KP causal
 vectorizer route.  Passing opens only a separately frozen short task gate;
 capture alone cannot raise the reviewer score.
 
+## Audited outcome (2026-08-06)
+
+V6-1 fails on clean source revision `98bfda7`.  The run executes the exact
+frozen 19 stages, 380 edge events, 760 batch-loss queries, and 48,640
+per-example observations.  All forward weights, readout weights, BatchNorm
+state/affines, and forward optimizer tensors remain bitwise fixed; the local
+fits use zero forward-weight reads and zero reverse-mode learning operations.
+
+The readout stage succeeds.  Its final feedback/forward parameter cosine is
+`0.81319`, and final-hidden teaching alignment is `0.87219`, compared with
+`0.00377` and `0.07787` under fixed HFA.  This confirms the post-failure V5
+diagnosis: per-example causal observations have enough signal to identify the
+small dense feedback map when fitted appropriately.
+
+The convolutional diagonal approximation fails.  Feedback norm ratios already
+reach `10.17--49.08` on finite late edges.  The predicted parent-field MSE
+grows from `0.0258` at edge 18 to `578` at edge 16, `3.12e7` at edge 15, and
+`6.84e30` at edge 9.  Edge 9 and all earlier stages become nonfinite.  The
+child contexts are strongly correlated, so treating their covariance as
+diagonal creates a multiplicative error cascade rather than a stable local
+inverse.
+
+Consequently every alignment and norm gate fails, no task-accuracy run is
+opened, and validation/test endpoints remain untouched.  Per the frozen rule,
+there is no ridge, clipping, or full-covariance recovery.  V6 closes the
+present task-loss-only no-KP causal vectorizer route.  Recovering convolutional
+feedback would require a materially different information source or an
+expensive full local system-identification method, which would need to beat
+the already measured KP, response-mirror, and BP cost/accuracy baselines.
diff --git a/RESULTS.md b/RESULTS.md
index 9bf803e..6cd5166 100644
--- a/RESULTS.md
+++ b/RESULTS.md
@@ -1147,6 +1147,20 @@ observations are informative near the output, not evidence that the no-KP
 vectorizer scales; the existing positive ResNet scaling result still belongs
 to the inherited KP substrate.
 
+The final bounded no-KP follow-up, Oral-A-v6, tests whether strict stagewise
+timing and locally whitened fits repair V5 without increasing its 760-query
+budget.  The dense readout fit works: feedback/forward cosine reaches
+`0.81319` and last-hidden teaching alignment reaches `0.87219`.  The
+convolutional diagonal fit does not.  Correlated child contexts make the
+diagonal inverse unstable: finite feedback norm ratios reach `49.08`, parent
+field MSE grows from `0.0258` at edge 18 to `6.84e30` at edge 9, and edge 9
+through the input become nonfinite.  All forward and BatchNorm state remains
+bitwise fixed, and no validation or test endpoint is evaluated.  This closes
+the present task-loss-only no-KP causal-vectorizer route.  A full covariance
+solver or local response measurement would be system identification; the
+latter is already represented by the response-mirror baseline, while KP is
+the substrate behind the positive standard-depth scaling result.
+
 ## How to run
 `experiments/run.py --mode {bp,fa,dfa,sdil} --dataset {mnist,fmnist,cifar10} --depth D --residual {0,1} --act {tanh,gelu,silu,relu}`
 Batteries: `experiments/run_v2.sh <ds> "<depths>" <res> <act> "<seeds>" <ep> <pfx>`.
diff --git a/results/oral_a_v6_calibration/result.json b/results/oral_a_v6_calibration/result.json
new file mode 100644
index 0000000..223f8ff
--- /dev/null
+++ b/results/oral_a_v6_calibration/result.json
@@ -0,0 +1,529 @@
+{
+  "architecture": {
+    "adaptive_feedback_parameters": 267904,
+    "family": "CIFAR 6n+2 ResNet, option-A shortcuts",
+    "feedback_macs_per_example": 40108672,
+    "forward_macs_per_example": 40551040,
+    "forward_parameters": 269722
+  },
+  "finite": false,
+  "fixed_hfa": {
+    "all_layer_alignment": 0.011278868932430143,
+    "early_third_alignment": -0.0002631605093483813,
+    "feedback_forward_cosine": [
+      0.0008809715509414673,
+      -0.002320552244782448,
+      0.03730485588312149,
+      0.02497659996151924,
+      -0.03438456356525421,
+      -0.001257734838873148,
+      0.023045765236020088,
+      0.010684739798307419,
+      -0.013966654427349567,
+      -0.03158111870288849,
+      0.009583890438079834,
+      -9.433389641344547e-05,
+      0.004490766208618879,
+      -0.004784357734024525,
+      0.004104882013052702,
+      0.004543168004602194,
+      -0.005818712059408426,
+      0.0014950999757274985,
+      0.003767170011997223
+    ],
+    "feedback_forward_norm_ratio": [
+      0.9996462464332581,
+      1.0550107955932617,
+      0.9593725800514221,
+      1.0017613172531128,
+      0.999921441078186,
+      0.9631232619285583,
+      0.9992203712463379,
+      0.9976226687431335,
+      1.0018895864486694,
+      1.0030434131622314,
+      1.0180330276489258,
+      1.0107296705245972,
+      0.9962885975837708,
+      0.9891077876091003,
+      1.0009011030197144,
+      1.0021207332611084,
+      1.0043002367019653,
+      1.0021206140518188,
+      0.9992877244949341
+    ],
+    "max_feedback_forward_norm_ratio": 1.0550107955932617,
+    "mean_feedback_forward_cosine": 0.0016142042954207252,
+    "min_feedback_forward_norm_ratio": 0.9593725800514221,
+    "per_layer": [
+      -0.0005718570901080966,
+      -0.0006265088450163603,
+      0.0012021202128380537,
+      3.7642239476554096e-05,
+      0.00048483084538020194,
+      -0.0021051904186606407,
+      0.00329246511682868,
+      0.00027297029737383127,
+      0.007543372455984354,
+      -0.0053173573687672615,
+      0.009830434806644917,
+      -0.0010977087076753378,
+      0.02368866838514805,
+      0.007433340419083834,
+      0.04359162598848343,
+      0.0057456037029623985,
+      0.039965368807315826,
+      0.0030560996383428574,
+      0.07787258923053741
+    ]
+  },
+  "hardware": {
+    "cuda_device_name": "NVIDIA GeForce GTX 1080",
+    "cuda_visible_devices": "5",
+    "device": "cuda",
+    "peak_memory_allocated_bytes": 1430980096,
+    "torch_version": "2.3.1+cu118"
+  },
+  "learned_scib": {
+    "all_layer_alignment": NaN,
+    "early_third_alignment": NaN,
+    "feedback_forward_cosine": [
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      -0.006127675995230675,
+      -0.011246882379055023,
+      0.021258916705846786,
+      0.00929059274494648,
+      -0.001609913888387382,
+      -0.0008538670372217894,
+      0.012138523161411285,
+      0.002846281975507736,
+      0.052370935678482056,
+      0.8131900429725647
+    ],
+    "feedback_forward_norm_ratio": [
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      7.744247913360596,
+      7.140369415283203,
+      11.147778511047363,
+      32.04150390625,
+      49.08123016357422,
+      48.10276412963867,
+      33.58113479614258,
+      10.170685768127441,
+      12.821575164794922,
+      1.120253562927246
+    ],
+    "max_feedback_forward_norm_ratio": NaN,
+    "mean_feedback_forward_cosine": NaN,
+    "min_feedback_forward_norm_ratio": NaN,
+    "per_layer": [
+      NaN,
+      0.0,
+      NaN,
+      0.0,
+      NaN,
+      0.0,
+      NaN,
+      0.0,
+      0.0,
+      -0.018046237528324127,
+      0.009331459179520607,
+      0.006288703065365553,
+      -0.0008923492860049009,
+      0.024693015962839127,
+      0.04579439014196396,
+      0.011086630634963512,
+      -0.025804810225963593,
+      0.3799460828304291,
+      0.8721908926963806
+    ]
+  },
+  "method_audit": {
+    "causal_query_normalization_state": "evaluation_running_statistics",
+    "forward_state_max_absolute_difference": 0.0,
+    "forward_weight_reads_in_feedback_fit": 0,
+    "ordinary_task_normalization_state": "not_run_forward_frozen",
+    "reverse_mode_learning_operations": 0,
+    "stage_order": [
+      "readout",
+      18,
+      17,
+      16,
+      15,
+      14,
+      13,
+      12,
+      11,
+      10,
+      9,
+      8,
+      7,
+      6,
+      5,
+      4,
+      3,
+      2,
+      1
+    ]
+  },
+  "protocol": "oral_a_v6_stagewise_whitened_causal_capture_v1",
+  "provenance": {
+    "git_commit": "98bfda74f7ee0457799b86465c867b6943c7be91",
+    "git_tracked_dirty": false
+  },
+  "schema_version": 1,
+  "settings": {
+    "alignment_probe": 64,
+    "batch_size": 128,
+    "calibration_augmentation": false,
+    "conv_diagonal_relative_ridge": 0.001,
+    "depth": 20,
+    "events_per_stage": 20,
+    "feedback_scale": 1.0,
+    "loader_seed": 0,
+    "normalization": "batchnorm",
+    "perturb_seed": 5000,
+    "readout_relative_ridge": 1e-06,
+    "residual_scale": 1.0,
+    "seed": 0,
+    "sigma": 0.01,
+    "split_seed": 2027,
+    "train_limit": 10000,
+    "val_examples": 5000,
+    "width": 16
+  },
+  "split": {
+    "dataset": "cifar10",
+    "input_layout": "NCHW",
+    "input_shape": [
+      3,
+      32,
+      32
+    ],
+    "loader_seed": 0,
+    "normalization_mean": [
+      0.49140000343322754,
+      0.4821999967098236,
+      0.4465000033378601
+    ],
+    "normalization_std": [
+      0.24699999392032623,
+      0.2434999942779541,
+      0.26159998774528503
+    ],
+    "split_from_training_only": true,
+    "split_seed": 2027,
+    "test_examples": 10000,
+    "train_examples": 10000,
+    "training_augmentation": "none",
+    "validation_class_counts": {
+      "0": 500,
+      "1": 500,
+      "2": 500,
+      "3": 500,
+      "4": 500,
+      "5": 500,
+      "6": 500,
+      "7": 500,
+      "8": 500,
+      "9": 500
+    },
+    "validation_examples": 5000,
+    "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b"
+  },
+  "stage_fits": [
+    {
+      "absolute_ridge": 0.00039648901367187497,
+      "after_mse": 1.4543522596359253,
+      "before_mse": 1.5032594203948975,
+      "examples": 2560,
+      "fit_target_cosine": 0.14180169892473388,
+      "kind": "readout",
+      "parameter_rms": 0.13902018685545237,
+      "relative_ridge": 1e-06
+    },
+    {
+      "absolute_ridge": 0.0007461613416671753,
+      "before_field_mse": 0.025809165835380555,
+      "edge": 18,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": 0.7461613416671753,
+      "minimum_diagonal": 0.0009920651791617274,
+      "observations": 20,
+      "parameter_rms": 0.753869841492034,
+      "parameter_update_rms": 0.756108481982398,
+      "relative_ridge": 0.001,
+      "target_rms": 0.16062293037888375
+    },
+    {
+      "absolute_ridge": 4.24671240234375,
+      "before_field_mse": 0.08534659817814827,
+      "edge": 17,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": 4246.71240234375,
+      "minimum_diagonal": 0.2649068832397461,
+      "observations": 20,
+      "parameter_rms": 0.6007491967588365,
+      "parameter_update_rms": 0.6027629071724261,
+      "relative_ridge": 0.001,
+      "target_rms": 0.20353771929970688
+    },
+    {
+      "absolute_ridge": 38247.564,
+      "before_field_mse": 578.0022918701172,
+      "edge": 16,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": 38247564.0,
+      "minimum_diagonal": 0.0,
+      "observations": 20,
+      "parameter_rms": 1.9794660509112565,
+      "parameter_update_rms": 1.9810645055729732,
+      "relative_ridge": 0.001,
+      "target_rms": 0.21561053942828906
+    },
+    {
+      "absolute_ridge": 2282466050.048,
+      "before_field_mse": 31215376.8,
+      "edge": 15,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": 2282466050048.0,
+      "minimum_diagonal": 0.0,
+      "observations": 20,
+      "parameter_rms": 2.828870306284442,
+      "parameter_update_rms": 2.8293681916124838,
+      "relative_ridge": 0.001,
+      "target_rms": 0.2689296525317363
+    },
+    {
+      "absolute_ridge": 147347425844002.8,
+      "before_field_mse": 1678965394636.8,
+      "edge": 14,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": 1.473474258440028e+17,
+      "minimum_diagonal": 2691794468864.0,
+      "observations": 20,
+      "parameter_rms": 2.899687546307522,
+      "parameter_update_rms": 2.9000687031993513,
+      "relative_ridge": 0.001,
+      "target_rms": 0.2908719252061969
+    },
+    {
+      "absolute_ridge": 9.83626578424372e+18,
+      "before_field_mse": 6.381369873326408e+16,
+      "edge": 13,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": 9.83626578424372e+21,
+      "minimum_diagonal": 0.0,
+      "observations": 20,
+      "parameter_rms": 2.68933867610627,
+      "parameter_update_rms": 2.6908325115291993,
+      "relative_ridge": 0.001,
+      "target_rms": 0.3458967232513598
+    },
+    {
+      "absolute_ridge": 2.310925825075764e+23,
+      "before_field_mse": 6.534132313707218e+20,
+      "edge": 12,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": 2.310925825075764e+26,
+      "minimum_diagonal": 4.643060183101878e+24,
+      "observations": 20,
+      "parameter_rms": 0.9222431094480169,
+      "parameter_update_rms": 0.9257461927512325,
+      "relative_ridge": 0.001,
+      "target_rms": 0.4134438229204942
+    },
+    {
+      "absolute_ridge": 5.315842145923256e+26,
+      "before_field_mse": 1.8040572061421976e+24,
+      "edge": 11,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": 5.315842145923256e+29,
+      "minimum_diagonal": 1.9429183075430184e+25,
+      "observations": 20,
+      "parameter_rms": 0.5912122703111267,
+      "parameter_update_rms": 0.596288099891043,
+      "relative_ridge": 0.001,
+      "target_rms": 0.5089343743931583
+    },
+    {
+      "absolute_ridge": 8.928898632191432e+29,
+      "before_field_mse": 2.5973411227762966e+27,
+      "edge": 10,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": 8.928898632191432e+32,
+      "minimum_diagonal": 1.7739228624702963e+31,
+      "observations": 20,
+      "parameter_rms": 0.6434051617818786,
+      "parameter_update_rms": 0.6484778518071089,
+      "relative_ridge": 0.001,
+      "target_rms": 0.5395306751768495
+    },
+    {
+      "absolute_ridge": NaN,
+      "before_field_mse": 6.838150174412349e+30,
+      "edge": 9,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": NaN,
+      "minimum_diagonal": NaN,
+      "observations": 20,
+      "parameter_rms": NaN,
+      "parameter_update_rms": NaN,
+      "relative_ridge": 0.001,
+      "target_rms": 0.6844558032335791
+    },
+    {
+      "absolute_ridge": NaN,
+      "before_field_mse": Infinity,
+      "edge": 8,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": NaN,
+      "minimum_diagonal": NaN,
+      "observations": 20,
+      "parameter_rms": NaN,
+      "parameter_update_rms": NaN,
+      "relative_ridge": 0.001,
+      "target_rms": 0.6733447699197179
+    },
+    {
+      "absolute_ridge": NaN,
+      "before_field_mse": NaN,
+      "edge": 7,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": NaN,
+      "minimum_diagonal": NaN,
+      "observations": 20,
+      "parameter_rms": NaN,
+      "parameter_update_rms": NaN,
+      "relative_ridge": 0.001,
+      "target_rms": 0.7612376527771475
+    },
+    {
+      "absolute_ridge": NaN,
+      "before_field_mse": Infinity,
+      "edge": 6,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": NaN,
+      "minimum_diagonal": NaN,
+      "observations": 20,
+      "parameter_rms": NaN,
+      "parameter_update_rms": NaN,
+      "relative_ridge": 0.001,
+      "target_rms": 0.9903483684468577
+    },
+    {
+      "absolute_ridge": NaN,
+      "before_field_mse": NaN,
+      "edge": 5,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": NaN,
+      "minimum_diagonal": NaN,
+      "observations": 20,
+      "parameter_rms": NaN,
+      "parameter_update_rms": NaN,
+      "relative_ridge": 0.001,
+      "target_rms": 1.2384468689881056
+    },
+    {
+      "absolute_ridge": NaN,
+      "before_field_mse": Infinity,
+      "edge": 4,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": NaN,
+      "minimum_diagonal": NaN,
+      "observations": 20,
+      "parameter_rms": NaN,
+      "parameter_update_rms": NaN,
+      "relative_ridge": 0.001,
+      "target_rms": 1.376840848200482
+    },
+    {
+      "absolute_ridge": NaN,
+      "before_field_mse": NaN,
+      "edge": 3,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": NaN,
+      "minimum_diagonal": NaN,
+      "observations": 20,
+      "parameter_rms": NaN,
+      "parameter_update_rms": NaN,
+      "relative_ridge": 0.001,
+      "target_rms": 1.7979800732137456
+    },
+    {
+      "absolute_ridge": NaN,
+      "before_field_mse": Infinity,
+      "edge": 2,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": NaN,
+      "minimum_diagonal": NaN,
+      "observations": 20,
+      "parameter_rms": NaN,
+      "parameter_update_rms": NaN,
+      "relative_ridge": 0.001,
+      "target_rms": 1.845518153372885
+    },
+    {
+      "absolute_ridge": NaN,
+      "before_field_mse": NaN,
+      "edge": 1,
+      "examples": 2560,
+      "kind": "convolution",
+      "mean_diagonal": NaN,
+      "minimum_diagonal": NaN,
+      "observations": 20,
+      "parameter_rms": NaN,
+      "parameter_update_rms": NaN,
+      "relative_ridge": 0.001,
+      "target_rms": 2.2282601185794344
+    }
+  ],
+  "test_examples_touched": 0,
+  "validation_endpoints_observed": 0,
+  "wall_seconds": 17.61427068710327,
+  "work": {
+    "clean_forward_examples": 48640,
+    "edge_events": 380,
+    "feedback_fit_macs_conservative_estimate": 5852657418240,
+    "forward_macs": 5917207756800,
+    "logical_batch_loss_queries": 760,
+    "per_example_causal_observations": 48640,
+    "per_example_cross_entropy_terms": 97280,
+    "perturbation_forward_examples": 97280,
+    "stages": 19,
+    "total_macs_conservative_estimate": 11769865175040
+  }
+}
diff --git a/results/oral_a_v6_calibration_gate.json b/results/oral_a_v6_calibration_gate.json
new file mode 100644
index 0000000..08bd3fa
--- /dev/null
+++ b/results/oral_a_v6_calibration_gate.json
@@ -0,0 +1,178 @@
+{
+  "checks": {
+    "all_layer_at_least_0.20": false,
+    "early_gain_over_fixed_hfa_at_least_0.08": false,
+    "early_third_at_least_0.10": false,
+    "exactly_19_stages": true,
+    "exactly_380_edge_events": true,
+    "exactly_48640_per_example_observations": true,
+    "exactly_760_batch_loss_queries": true,
+    "feedback_norm_ratios_in_0.1_to_3": false,
+    "finite": false,
+    "forward_state_bitwise_fixed": true,
+    "stage_order_readout_then_18_to_1": true,
+    "zero_forward_weight_reads_in_fit": true,
+    "zero_reverse_mode_learning_operations": true
+  },
+  "conditional_short_task_gate_open": false,
+  "confirmation_test_seeds_touched": false,
+  "fixed_hfa": {
+    "all_layer_alignment": 0.011278868932430143,
+    "early_third_alignment": -0.0002631605093483813,
+    "feedback_forward_cosine": [
+      0.0008809715509414673,
+      -0.002320552244782448,
+      0.03730485588312149,
+      0.02497659996151924,
+      -0.03438456356525421,
+      -0.001257734838873148,
+      0.023045765236020088,
+      0.010684739798307419,
+      -0.013966654427349567,
+      -0.03158111870288849,
+      0.009583890438079834,
+      -9.433389641344547e-05,
+      0.004490766208618879,
+      -0.004784357734024525,
+      0.004104882013052702,
+      0.004543168004602194,
+      -0.005818712059408426,
+      0.0014950999757274985,
+      0.003767170011997223
+    ],
+    "feedback_forward_norm_ratio": [
+      0.9996462464332581,
+      1.0550107955932617,
+      0.9593725800514221,
+      1.0017613172531128,
+      0.999921441078186,
+      0.9631232619285583,
+      0.9992203712463379,
+      0.9976226687431335,
+      1.0018895864486694,
+      1.0030434131622314,
+      1.0180330276489258,
+      1.0107296705245972,
+      0.9962885975837708,
+      0.9891077876091003,
+      1.0009011030197144,
+      1.0021207332611084,
+      1.0043002367019653,
+      1.0021206140518188,
+      0.9992877244949341
+    ],
+    "max_feedback_forward_norm_ratio": 1.0550107955932617,
+    "mean_feedback_forward_cosine": 0.0016142042954207252,
+    "min_feedback_forward_norm_ratio": 0.9593725800514221,
+    "per_layer": [
+      -0.0005718570901080966,
+      -0.0006265088450163603,
+      0.0012021202128380537,
+      3.7642239476554096e-05,
+      0.00048483084538020194,
+      -0.0021051904186606407,
+      0.00329246511682868,
+      0.00027297029737383127,
+      0.007543372455984354,
+      -0.0053173573687672615,
+      0.009830434806644917,
+      -0.0010977087076753378,
+      0.02368866838514805,
+      0.007433340419083834,
+      0.04359162598848343,
+      0.0057456037029623985,
+      0.039965368807315826,
+      0.0030560996383428574,
+      0.07787258923053741
+    ]
+  },
+  "learned_scib": {
+    "all_layer_alignment": NaN,
+    "early_third_alignment": NaN,
+    "feedback_forward_cosine": [
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      -0.006127675995230675,
+      -0.011246882379055023,
+      0.021258916705846786,
+      0.00929059274494648,
+      -0.001609913888387382,
+      -0.0008538670372217894,
+      0.012138523161411285,
+      0.002846281975507736,
+      0.052370935678482056,
+      0.8131900429725647
+    ],
+    "feedback_forward_norm_ratio": [
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      NaN,
+      7.744247913360596,
+      7.140369415283203,
+      11.147778511047363,
+      32.04150390625,
+      49.08123016357422,
+      48.10276412963867,
+      33.58113479614258,
+      10.170685768127441,
+      12.821575164794922,
+      1.120253562927246
+    ],
+    "max_feedback_forward_norm_ratio": NaN,
+    "mean_feedback_forward_cosine": NaN,
+    "min_feedback_forward_norm_ratio": NaN,
+    "per_layer": [
+      NaN,
+      0.0,
+      NaN,
+      0.0,
+      NaN,
+      0.0,
+      NaN,
+      0.0,
+      0.0,
+      -0.018046237528324127,
+      0.009331459179520607,
+      0.006288703065365553,
+      -0.0008923492860049009,
+      0.024693015962839127,
+      0.04579439014196396,
+      0.011086630634963512,
+      -0.025804810225963593,
+      0.3799460828304291,
+      0.8721908926963806
+    ]
+  },
+  "protocol": "oral_a_v6_stagewise_whitened_causal_capture_gate_v1",
+  "review_score_after": 5,
+  "review_score_before": 5,
+  "score_change_rule": "causal capture alone cannot raise score",
+  "source_commit": "98bfda74f7ee0457799b86465c867b6943c7be91",
+  "source_result": "results/oral_a_v6_calibration/result.json",
+  "status": "failed",
+  "work": {
+    "clean_forward_examples": 48640,
+    "edge_events": 380,
+    "feedback_fit_macs_conservative_estimate": 5852657418240,
+    "forward_macs": 5917207756800,
+    "logical_batch_loss_queries": 760,
+    "per_example_causal_observations": 48640,
+    "per_example_cross_entropy_terms": 97280,
+    "perturbation_forward_examples": 97280,
+    "stages": 19,
+    "total_macs_conservative_estimate": 11769865175040
+  }
+}