protocol: freeze calibration-split oral-B-v2 recovery

7 files changed, 1244 insertions, 0 deletions

diff --git a/experiments/analyze_bci_v2_calibrated_confirmation.py b/experiments/analyze_bci_v2_calibrated_confirmation.py
new file mode 100644
index 0000000..9fd5eaf
--- /dev/null
+++ b/experiments/analyze_bci_v2_calibrated_confirmation.py
@@ -0,0 +1,363 @@
+#!/usr/bin/env python3
+"""Audit untouched calibration-split oral-B-v2 confirmation."""
+import argparse
+import glob
+import hashlib
+import json
+import math
+import os
+
+from experiments.analyze_bci_v2_recovery_confirmation import (
+    checks,
+    summarize,
+    task_clusters,
+)
+from experiments.bci_v2_calibrated_confirmation import (
+    MODEL_SEEDS,
+    TASK_SEEDS,
+)
+from experiments.bci_v2_calibrated_run import (
+    CALIBRATION_EPISODES,
+    CHALLENGE_EPISODES,
+    TARGET_QUANTILES,
+    build_recovery_config,
+    source_paths,
+)
+from experiments.bci_v2_recovery_run import FIXED_CONFIG
+from experiments.bci_v2_run import CONDITIONS
+
+
+ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+RUNNER_PATH = os.path.join(
+    ROOT, "experiments", "bci_v2_calibrated_confirmation.py"
+)
+
+
+def require(condition, message):
+    if not condition:
+        raise ValueError(message)
+
+
+def sha256(path):
+    digest = hashlib.sha256()
+    with open(path, "rb") as handle:
+        for chunk in iter(lambda: handle.read(1024 * 1024), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def finite_tree(value):
+    if isinstance(value, dict):
+        return all(finite_tree(item) for item in value.values())
+    if isinstance(value, (list, tuple)):
+        return all(finite_tree(item) for item in value)
+    if isinstance(value, (int, float)):
+        return math.isfinite(value)
+    return True
+
+
+def confirmation_paths(r1_gate):
+    paths = source_paths()
+    paths["development_runner"] = paths["runner"]
+    paths["runner"] = RUNNER_PATH
+    paths["r1_gate"] = os.path.abspath(r1_gate)
+    return paths
+
+
+def validate(row, path, digests):
+    require(row.get("schema_version") == 4, f"{path}: schema")
+    args = row.get("args", {})
+    task_seed = args.get("task_seed")
+    model_seed = args.get("model_seed")
+    require(
+        task_seed in TASK_SEEDS and model_seed in MODEL_SEEDS,
+        f"{path}: seeds",
+    )
+    protocol = row.get("protocol", {})
+    require(
+        protocol.get("name")
+        == "oral_b_v2_calibrated_recovery_confirmation_v1"
+        and protocol.get("split") == "untouched_confirmation"
+        and protocol.get("training_task_seed") == task_seed
+        and protocol.get("model_seed") == model_seed
+        and protocol.get("fixed_config") == FIXED_CONFIG
+        and protocol.get("calibration_uses_outcome_labels") is False
+        and protocol.get("no_further_selection") is True
+        and protocol.get("confirmation_grid_size") == 30
+        and protocol.get("protocol_sha256") == digests["protocol"]
+        and protocol.get("r1_gate_sha256") == digests["r1_gate"],
+        f"{path}: protocol",
+    )
+    provenance = row.get("provenance", {})
+    require(
+        provenance.get("git_tracked_dirty") is False
+        and provenance.get("tracked_inputs")
+        and all(provenance["tracked_inputs"].values())
+        and set(provenance["tracked_inputs"]) == set(digests)
+        and provenance.get("input_sha256") == digests,
+        f"{path}: provenance",
+    )
+    require(
+        row.get("config") == vars(build_recovery_config())
+        and row.get("split") == "untouched_confirmation"
+        and row.get("finite") is True
+        and finite_tree(row),
+        f"{path}: finite/config",
+    )
+    require(
+        set(row.get("conditions", {})) == set(CONDITIONS),
+        f"{path}: conditions",
+    )
+    for name in CONDITIONS:
+        warmup = row["warmup"][name]
+        condition = row["conditions"][name]
+        cost = condition["cost"]
+        require(
+            warmup["batches"] == 100
+            and warmup["examples"] == 6400
+            and warmup["instruction_present"] is False
+            and warmup["role_cursor_scalar_observations"] == 12800
+            and warmup["predictor_max_abs_error"] <= 1e-5
+            and len(condition["daily_success"]) == 14
+            and cost["maximum_state_episode_steps"] == 25088
+            and cost["cursor_scalar_observations"]
+            == 2 * cost["role_probe_examples"]
+            and cost["task_loss_queries"] == 0
+            and cost["reverse_mode_calls"] == 0,
+            f"{path}: condition {name}",
+        )
+    challenge = row["assays"]["challenge"]
+    calibration = challenge["calibration"]
+    targets = challenge["targets"]
+    require(
+        calibration["episodes"] == CALIBRATION_EPISODES
+        and calibration["quantiles"] == list(TARGET_QUANTILES)
+        and calibration["targets"] == targets
+        and calibration["uses_outcome_labels"] is False
+        and calibration["seed"] == 600_000 + task_seed
+        and challenge["episodes_per_target"] == CHALLENGE_EPISODES
+        and challenge["selection_over_evaluation_outcomes"] is False
+        and challenge["maximum_steps_per_episode"] == 28
+        and len(targets) == len(TARGET_QUANTILES)
+        and all(
+            targets[index] < targets[index + 1]
+            for index in range(len(targets) - 1)
+        )
+        and row["signatures"]["challenge_episodes"]
+        == len(TARGET_QUANTILES) * CHALLENGE_EPISODES,
+        f"{path}: calibration",
+    )
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--results",
+        default="results/bci_v2_calibrated_confirmation",
+    )
+    parser.add_argument(
+        "--r1-gate",
+        default="results/bci_v2_calibrated_dev_gate.json",
+    )
+    parser.add_argument(
+        "--out",
+        default="results/bci_v2_calibrated_confirmation_gate.json",
+    )
+    args = parser.parse_args()
+    with open(args.r1_gate) as handle:
+        r1 = json.load(handle)
+    require(
+        r1.get("protocol")
+        == "oral_b_v2_calibrated_recovery_development_v1"
+        and r1.get("status") == "passed"
+        and r1.get("complete_grid") is True
+        and r1.get("fixed_config") == FIXED_CONFIG
+        and r1.get("confirmation_seeds_touched") is False
+        and r1.get("calibrated_confirmation_opened") is True,
+        "calibrated R1 gate",
+    )
+    development_digests = {
+        name: sha256(path)
+        for name, path in source_paths().items()
+    }
+    require(
+        r1.get("input_sha256") == development_digests,
+        "R1 source binding",
+    )
+    digests = {
+        name: sha256(path)
+        for name, path in confirmation_paths(args.r1_gate).items()
+    }
+    expected = {
+        f"bci_v2_calibrated_confirm_t{task_seed}_m{model_seed}.json"
+        for task_seed in TASK_SEEDS
+        for model_seed in MODEL_SEEDS
+    }
+    observed = {
+        os.path.basename(path)
+        for path in glob.glob(os.path.join(args.results, "*.json"))
+    }
+    require(
+        observed == expected,
+        (
+            f"confirmation drift: missing={sorted(expected-observed)}, "
+            f"extra={sorted(observed-expected)}"
+        ),
+    )
+    records = {}
+    commits = set()
+    source_sha256 = {}
+    for task_seed in TASK_SEEDS:
+        for model_seed in MODEL_SEEDS:
+            path = os.path.join(
+                args.results,
+                (
+                    f"bci_v2_calibrated_confirm_t{task_seed}"
+                    f"_m{model_seed}.json"
+                ),
+            )
+            with open(path) as handle:
+                row = json.load(handle)
+            validate(row, path, digests)
+            records[(task_seed, model_seed)] = row
+            commits.add(row["provenance"]["git_commit"])
+            source_sha256[path] = sha256(path)
+    require(len(commits) == 1, "confirmation source revision drift")
+
+    getters = {
+        "intact_final": lambda row:
+            row["conditions"]["intact"]["final_success"],
+        "intact_gain": lambda row:
+            row["conditions"]["intact"]["learning_gain"],
+        "fixed_role_gap": lambda row: (
+            row["conditions"]["intact"]["final_success"]
+            - row["conditions"]["fixed_role"]["final_success"]
+        ),
+        "oracle_deficit": lambda row: (
+            row["conditions"]["oracle_role"]["final_success"]
+            - row["conditions"]["intact"]["final_success"]
+        ),
+        "plasticity_half_margin": lambda row: (
+            0.5 * row["conditions"]["intact"]["learning_gain"]
+            - row["conditions"]["plasticity_lesion"]["learning_gain"]
+        ),
+        "role_cosine": lambda row:
+            row["conditions"]["intact"]["role_cosine_after_training"],
+        "critic_training_gap": lambda row: (
+            row["conditions"]["intact"]["final_success"]
+            - row["conditions"]["critic_training_lesion"][
+                "final_success"
+            ]
+        ),
+        "outcome_training_gap": lambda row: (
+            row["conditions"]["intact"]["final_success"]
+            - row["conditions"]["outcome_training_lesion"][
+                "final_success"
+            ]
+        ),
+        "residual_soma_corr": lambda row:
+            row["signatures"]["mean_abs_residual_soma_corr"],
+        "raw_residual_corr_gap": lambda row:
+            row["signatures"][
+                "raw_minus_residual_abs_soma_corr"
+            ],
+        "surrounding_accuracy": lambda row:
+            row["signatures"][
+                "surrounding_event_decoder_balanced_acc"
+            ],
+        "decoder_corr": lambda row:
+            row["signatures"]["decoder_distance_residual_corr"],
+        "sign_inversion": lambda row:
+            row["signatures"][
+                "causal_role_sign_inversion_index"
+            ],
+        "velocity_advantage": lambda row:
+            row["signatures"][
+                "velocity_minus_error_abs_cv_corr"
+            ],
+        "challenge_fraction": lambda row:
+            row["signatures"]["challenge_success_fraction"],
+        "terminal_accuracy": lambda row:
+            row["signatures"][
+                "terminal_residual_outcome_balanced_acc"
+            ],
+        "terminal_separation": lambda row:
+            row["signatures"][
+                "terminal_role_aligned_outcome_separation"
+            ],
+        "outcome_lesion_drop": lambda row:
+            row["signatures"][
+                "terminal_outcome_separation_drop_under_acute_lesion"
+            ],
+        "critic_expectedness": lambda row:
+            row["signatures"][
+                "mean_critic_expectedness_contribution"
+            ],
+        "critic_value_corr": lambda row:
+            row["signatures"][
+                "critic_contribution_value_prediction_corr"
+            ],
+    }
+    clustered = {
+        name: task_clusters(records, getter)
+        for name, getter in getters.items()
+    }
+    metrics = {
+        name: summarize(values)
+        for name, values in clustered.items()
+    }
+    all_values = {
+        name: [getter(row) for row in records.values()]
+        for name, getter in getters.items()
+    }
+    gate_checks = checks(metrics, clustered, all_values)
+    passed = all(
+        value
+        for category in gate_checks.values()
+        for value in (
+            [category]
+            if isinstance(category, bool)
+            else category.values()
+        )
+    )
+    output = {
+        "protocol":
+            "oral_b_v2_calibrated_recovery_confirmation_v1",
+        "status": "passed" if passed else "failed",
+        "complete_grid": True,
+        "fixed_config": FIXED_CONFIG,
+        "calibration_quantiles": list(TARGET_QUANTILES),
+        "checks": gate_checks,
+        "metrics": metrics,
+        "positive_sign_count": sum(
+            value > 0 for value in all_values["sign_inversion"]
+        ),
+        "source_commit": next(iter(commits)),
+        "source_sha256": source_sha256,
+        "input_sha256": digests,
+        "oral_b_v2_outcome_surprise_established": passed,
+        "all_prior_failures_preserved": True,
+        "old_oral_a_gate_remains_closed": True,
+        "new_oral_a_v2_protocol_may_be_frozen": passed,
+        "review_score_before": 7,
+        "review_score_after": 8 if passed else 7,
+        "score_change_rule": (
+            "only a complete untouched calibrated R2 pass establishes "
+            "role-vectorized TD outcome surprise; prior failures and "
+            "the old oral-A gate remain unchanged"
+        ),
+    }
+    os.makedirs(os.path.dirname(os.path.abspath(args.out)), exist_ok=True)
+    if os.path.exists(args.out):
+        with open(args.out) as handle:
+            existing = json.load(handle)
+        require(existing == output, "existing calibrated R2 gate differs")
+    else:
+        with open(args.out, "w") as handle:
+            json.dump(output, handle, indent=2, sort_keys=True)
+            handle.write("\n")
+    print(json.dumps(output, indent=2))
+
+
+if __name__ == "__main__":
+    main()
diff --git a/experiments/analyze_bci_v2_calibrated_development.py b/experiments/analyze_bci_v2_calibrated_development.py
new file mode 100644
index 0000000..e8781d6
--- /dev/null
+++ b/experiments/analyze_bci_v2_calibrated_development.py
@@ -0,0 +1,217 @@
+#!/usr/bin/env python3
+"""Audit calibration-split oral-B-v2 development validation."""
+import argparse
+import glob
+import hashlib
+import json
+import os
+
+from experiments.analyze_bci_v2_recovery_development import checks
+from experiments.bci_v2_calibrated_run import (
+    CALIBRATION_EPISODES,
+    CHALLENGE_EPISODES,
+    MODEL_SEEDS,
+    TARGET_QUANTILES,
+    TASK_SEEDS,
+    build_recovery_config,
+    source_paths,
+)
+from experiments.bci_v2_recovery_run import FIXED_CONFIG
+from experiments.bci_v2_run import CONDITIONS
+
+
+def require(condition, message):
+    if not condition:
+        raise ValueError(message)
+
+
+def sha256(path):
+    digest = hashlib.sha256()
+    with open(path, "rb") as handle:
+        for chunk in iter(lambda: handle.read(1024 * 1024), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def validate(row, path, task_seed, digests):
+    require(row.get("schema_version") == 4, f"{path}: schema")
+    require(
+        row.get("args", {}).get("task_seed") == task_seed
+        and row["args"].get("model_seed") in MODEL_SEEDS,
+        f"{path}: seeds",
+    )
+    protocol = row.get("protocol", {})
+    require(
+        protocol.get("name")
+        == "oral_b_v2_calibrated_recovery_development_v1"
+        and protocol.get("selection_split") == "development_validation"
+        and protocol.get("hyperparameter_selection") is False
+        and protocol.get("confirmation_seeds_touched") is False
+        and protocol.get("fixed_config") == FIXED_CONFIG
+        and protocol.get("calibration_uses_outcome_labels") is False
+        and protocol.get("protocol_sha256") == digests["protocol"]
+        and protocol.get("failed_target_gate_sha256")
+        == digests["failed_target_gate"],
+        f"{path}: protocol",
+    )
+    provenance = row.get("provenance", {})
+    require(
+        provenance.get("git_tracked_dirty") is False
+        and provenance.get("tracked_inputs")
+        and all(provenance["tracked_inputs"].values())
+        and set(provenance["tracked_inputs"]) == set(digests)
+        and provenance.get("input_sha256") == digests,
+        f"{path}: provenance",
+    )
+    require(
+        row.get("config") == vars(build_recovery_config())
+        and row.get("split") == "development_validation"
+        and row.get("finite") is True,
+        f"{path}: config/split",
+    )
+    require(
+        set(row.get("conditions", {})) == set(CONDITIONS),
+        f"{path}: conditions",
+    )
+    for name in CONDITIONS:
+        warmup = row["warmup"][name]
+        condition = row["conditions"][name]
+        cost = condition["cost"]
+        require(
+            warmup["batches"] == 100
+            and warmup["examples"] == 6400
+            and warmup["instruction_present"] is False
+            and warmup["role_cursor_scalar_observations"] == 12800
+            and warmup["predictor_max_abs_error"] <= 1e-5
+            and len(condition["daily_success"]) == 14
+            and cost["maximum_state_episode_steps"] == 25088
+            and cost["cursor_scalar_observations"]
+            == 2 * cost["role_probe_examples"]
+            and cost["task_loss_queries"] == 0
+            and cost["reverse_mode_calls"] == 0,
+            f"{path}: condition {name}",
+        )
+    challenge = row["assays"]["challenge"]
+    calibration = challenge["calibration"]
+    targets = challenge["targets"]
+    require(
+        calibration["episodes"] == CALIBRATION_EPISODES
+        and calibration["quantiles"] == list(TARGET_QUANTILES)
+        and calibration["targets"] == targets
+        and calibration["uses_outcome_labels"] is False
+        and calibration["seed"] == 550_000 + task_seed
+        and challenge["episodes_per_target"] == CHALLENGE_EPISODES
+        and challenge["selection_over_evaluation_outcomes"] is False
+        and challenge["maximum_steps_per_episode"] == 28
+        and len(targets) == len(TARGET_QUANTILES)
+        and all(
+            targets[index] < targets[index + 1]
+            for index in range(len(targets) - 1)
+        )
+        and row["signatures"]["challenge_episodes"]
+        == len(TARGET_QUANTILES) * CHALLENGE_EPISODES,
+        f"{path}: calibration",
+    )
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--results", default="results/bci_v2_calibrated_dev"
+    )
+    parser.add_argument(
+        "--out", default="results/bci_v2_calibrated_dev_gate.json"
+    )
+    args = parser.parse_args()
+    digests = {
+        name: sha256(path)
+        for name, path in source_paths().items()
+    }
+    expected = {
+        f"bci_v2_calibrated_t{task_seed}_m0.json"
+        for task_seed in TASK_SEEDS
+    }
+    observed = {
+        os.path.basename(path)
+        for path in glob.glob(os.path.join(args.results, "*.json"))
+    }
+    require(
+        observed == expected,
+        (
+            f"calibrated grid drift: missing={sorted(expected-observed)}, "
+            f"extra={sorted(observed-expected)}"
+        ),
+    )
+    records = {}
+    commits = set()
+    source_sha256 = {}
+    for task_seed in TASK_SEEDS:
+        path = os.path.join(
+            args.results,
+            f"bci_v2_calibrated_t{task_seed}_m0.json",
+        )
+        with open(path) as handle:
+            row = json.load(handle)
+        validate(row, path, task_seed, digests)
+        records[task_seed] = row
+        commits.add(row["provenance"]["git_commit"])
+        source_sha256[path] = sha256(path)
+    require(len(commits) == 1, "calibrated source revision drift")
+    checks_by_seed = {
+        str(seed): checks(records[seed]) for seed in TASK_SEEDS
+    }
+    passed = all(
+        value for seed_checks in checks_by_seed.values()
+        for value in seed_checks.values()
+    )
+    output = {
+        "protocol":
+            "oral_b_v2_calibrated_recovery_development_v1",
+        "status": "passed" if passed else "failed",
+        "complete_grid": True,
+        "fixed_config": FIXED_CONFIG,
+        "calibration_quantiles": list(TARGET_QUANTILES),
+        "checks_by_task_seed": checks_by_seed,
+        "calibrated_targets_by_task_seed": {
+            str(seed): records[seed]["assays"]["challenge"]["targets"]
+            for seed in TASK_SEEDS
+        },
+        "intact_final_by_task_seed": [
+            records[seed]["conditions"]["intact"]["final_success"]
+            for seed in TASK_SEEDS
+        ],
+        "challenge_success_fraction_by_task_seed": [
+            records[seed]["signatures"]["challenge_success_fraction"]
+            for seed in TASK_SEEDS
+        ],
+        "terminal_outcome_accuracy_by_task_seed": [
+            records[seed]["signatures"][
+                "terminal_residual_outcome_balanced_acc"
+            ]
+            for seed in TASK_SEEDS
+        ],
+        "confirmation_seeds_touched": False,
+        "calibrated_confirmation_opened": passed,
+        "source_commit": next(iter(commits)),
+        "source_sha256": source_sha256,
+        "input_sha256": digests,
+        "review_score_before": 7,
+        "review_score_after": 7,
+        "score_change_rule": (
+            "development validation never changes the formal score"
+        ),
+    }
+    os.makedirs(os.path.dirname(os.path.abspath(args.out)), exist_ok=True)
+    if os.path.exists(args.out):
+        with open(args.out) as handle:
+            existing = json.load(handle)
+        require(existing == output, "existing calibrated R1 gate differs")
+    else:
+        with open(args.out, "w") as handle:
+            json.dump(output, handle, indent=2, sort_keys=True)
+            handle.write("\n")
+    print(json.dumps(output, indent=2))
+
+
+if __name__ == "__main__":
+    main()
diff --git a/experiments/bci_v2_calibrated_confirmation.py b/experiments/bci_v2_calibrated_confirmation.py
new file mode 100644
index 0000000..ae79708
--- /dev/null
+++ b/experiments/bci_v2_calibrated_confirmation.py
@@ -0,0 +1,143 @@
+#!/usr/bin/env python3
+"""Run one untouched calibration-split oral-B-v2 confirmation cell."""
+import argparse
+import json
+import os
+import sys
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+from experiments.bci_v2_calibrated_run import (
+    PROTOCOL_PATH,
+    finite_tree,
+    provenance,
+    require_parent_gates,
+    run_cell,
+    sha256,
+    source_paths,
+)
+from experiments.bci_v2_recovery_run import FIXED_CONFIG
+
+
+ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+R1_GATE_PATH = os.path.join(
+    ROOT, "results", "bci_v2_calibrated_dev_gate.json"
+)
+TASK_SEEDS = tuple(range(30, 36))
+MODEL_SEEDS = tuple(range(5))
+
+
+def require_r1_gate(r1):
+    digests = {
+        name: sha256(path)
+        for name, path in source_paths().items()
+    }
+    if not (
+        r1.get("protocol")
+        == "oral_b_v2_calibrated_recovery_development_v1"
+        and r1.get("status") == "passed"
+        and r1.get("complete_grid") is True
+        and r1.get("fixed_config") == FIXED_CONFIG
+        and r1.get("confirmation_seeds_touched") is False
+        and r1.get("calibrated_confirmation_opened") is True
+        and r1.get("review_score_after") == 7
+        and r1.get("input_sha256") == digests
+    ):
+        raise ValueError(
+            "calibrated R2 requires the complete eligible R1 gate"
+        )
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--task-seed", type=int, choices=TASK_SEEDS, required=True
+    )
+    parser.add_argument(
+        "--model-seed", type=int, choices=MODEL_SEEDS, required=True
+    )
+    parser.add_argument(
+        "--r1-gate",
+        default="results/bci_v2_calibrated_dev_gate.json",
+    )
+    parser.add_argument(
+        "--outdir", default="results/bci_v2_calibrated_confirmation"
+    )
+    args = parser.parse_args()
+    require_parent_gates()
+    with open(args.r1_gate) as handle:
+        r1 = json.load(handle)
+    require_r1_gate(r1)
+    source = provenance({
+        "development_runner": os.path.join(
+            ROOT, "experiments", "bci_v2_calibrated_run.py"
+        ),
+        "runner": os.path.abspath(__file__),
+        "r1_gate": os.path.abspath(args.r1_gate),
+    })
+    if (
+        source["git_tracked_dirty"]
+        or not all(source["tracked_inputs"].values())
+    ):
+        raise RuntimeError(
+            "calibrated R2 requires clean, tracked, frozen inputs"
+        )
+    cell = run_cell(
+        args.task_seed,
+        args.model_seed,
+        split="untouched_confirmation",
+        performance_seed_offset=590_000,
+        calibration_seed_offset=600_000,
+        challenge_seed_offset=610_000,
+    )
+    result = {
+        "schema_version": 4,
+        "protocol": {
+            "name":
+                "oral_b_v2_calibrated_recovery_confirmation_v1",
+            "split": "untouched_confirmation",
+            "training_task_seed": args.task_seed,
+            "model_seed": args.model_seed,
+            "fixed_config": FIXED_CONFIG,
+            "calibration_uses_outcome_labels": False,
+            "no_further_selection": True,
+            "confirmation_grid_size": (
+                len(TASK_SEEDS) * len(MODEL_SEEDS)
+            ),
+            "protocol_sha256": sha256(PROTOCOL_PATH),
+            "r1_gate_sha256": sha256(args.r1_gate),
+        },
+        "args": vars(args),
+        "provenance": source,
+        **cell,
+    }
+    result["finite"] = finite_tree(result)
+    if not result["finite"]:
+        raise RuntimeError("non-finite calibrated confirmation record")
+    os.makedirs(args.outdir, exist_ok=True)
+    path = os.path.join(
+        args.outdir,
+        (
+            f"bci_v2_calibrated_confirm_t{args.task_seed}"
+            f"_m{args.model_seed}.json"
+        ),
+    )
+    if os.path.exists(path):
+        raise FileExistsError(f"refusing to overwrite {path}")
+    with open(path, "w") as handle:
+        json.dump(result, handle, indent=2, sort_keys=True)
+        handle.write("\n")
+    print(json.dumps({
+        "path": path,
+        "intact_final": result["conditions"]["intact"]["final_success"],
+        "challenge_success_fraction": result["signatures"][
+            "challenge_success_fraction"
+        ],
+        "terminal_outcome_accuracy": result["signatures"][
+            "terminal_residual_outcome_balanced_acc"
+        ],
+        "finite": result["finite"],
+    }, indent=2))
+
+
+if __name__ == "__main__":
+    main()
diff --git a/experiments/bci_v2_calibrated_confirmation.sh b/experiments/bci_v2_calibrated_confirmation.sh
new file mode 100644
index 0000000..73e8580
--- /dev/null
+++ b/experiments/bci_v2_calibrated_confirmation.sh
@@ -0,0 +1,16 @@
+#!/usr/bin/env bash
+# Frozen calibrated R2: 6 untouched task seeds x 5 model seeds.
+set -euo pipefail
+
+export OMP_NUM_THREADS=1
+export MKL_NUM_THREADS=1
+
+for task_seed in 30 31 32 33 34 35; do
+  for model_seed in 0 1 2 3 4; do
+    python experiments/bci_v2_calibrated_confirmation.py \
+      --task-seed "$task_seed" \
+      --model-seed "$model_seed"
+  done
+done
+
+PYTHONPATH=. python experiments/analyze_bci_v2_calibrated_confirmation.py
diff --git a/experiments/bci_v2_calibrated_development.sh b/experiments/bci_v2_calibrated_development.sh
new file mode 100644
index 0000000..0a9f004
--- /dev/null
+++ b/experiments/bci_v2_calibrated_development.sh
@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+# Frozen calibrated R1: one fixed mechanism x 3 new validation seeds.
+set -euo pipefail
+
+export OMP_NUM_THREADS=1
+export MKL_NUM_THREADS=1
+
+for task_seed in 26 27 28; do
+  python experiments/bci_v2_calibrated_run.py --task-seed "$task_seed"
+done
+
+PYTHONPATH=. python experiments/analyze_bci_v2_calibrated_development.py
diff --git a/experiments/bci_v2_calibrated_run.py b/experiments/bci_v2_calibrated_run.py
new file mode 100644
index 0000000..cb477dd
--- /dev/null
+++ b/experiments/bci_v2_calibrated_run.py
@@ -0,0 +1,457 @@
+#!/usr/bin/env python3
+"""Run one calibration-split oral-B-v2 development cell."""
+import argparse
+from dataclasses import replace
+import hashlib
+import json
+import os
+import platform
+import resource
+import subprocess
+import sys
+import time
+
+import torch
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+from experiments.bci_v2_recovery_run import (
+    FIXED_CONFIG,
+    build_recovery_config,
+)
+from experiments.bci_v2_run import (
+    ACUTE_MODES,
+    CONDITIONS,
+    evaluate_performance,
+    finite_tree,
+    neutral_warmup,
+    train,
+)
+from sdil.bci import generate_trajectories
+from sdil.bci_v2 import BCIV2, run_day_v2
+from sdil.bci_v2_recovery_metrics import (
+    annotate_target_events,
+    recovery_signature_metrics,
+)
+
+
+ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+PROTOCOL_PATH = os.path.join(
+    ROOT, "ORAL_B_V2_CALIBRATED_RECOVERY.md"
+)
+D4_GATE_PATH = os.path.join(
+    ROOT, "results", "kp_dynamic_projection_confirmation_gate.json"
+)
+OLD_R2_GATE_PATH = os.path.join(
+    ROOT, "results", "bci_td_confirmation_gate.json"
+)
+FAILED_V2_GATE_PATH = os.path.join(
+    ROOT, "results", "bci_v2_dev_gate.json"
+)
+FAILED_TARGET_GATE_PATH = os.path.join(
+    ROOT, "results", "bci_v2_recovery_dev_gate.json"
+)
+TASK_SEEDS = (26, 27, 28)
+MODEL_SEEDS = (0,)
+CALIBRATION_EPISODES = 512
+TARGET_QUANTILES = (0.20, 0.35, 0.50, 0.65, 0.80)
+CHALLENGE_EPISODES = 128
+UNCENSORED_TARGET = 2.1
+
+
+def sha256(path):
+    digest = hashlib.sha256()
+    with open(path, "rb") as handle:
+        for chunk in iter(lambda: handle.read(1024 * 1024), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def source_paths():
+    return {
+        "runner": os.path.abspath(__file__),
+        "development_analyzer": os.path.join(
+            ROOT,
+            "experiments",
+            "analyze_bci_v2_calibrated_development.py",
+        ),
+        "confirmation_runner": os.path.join(
+            ROOT, "experiments", "bci_v2_calibrated_confirmation.py"
+        ),
+        "confirmation_analyzer": os.path.join(
+            ROOT,
+            "experiments",
+            "analyze_bci_v2_calibrated_confirmation.py",
+        ),
+        "common_runner": os.path.join(
+            ROOT, "experiments", "bci_v2_run.py"
+        ),
+        "recovery_runner": os.path.join(
+            ROOT, "experiments", "bci_v2_recovery_run.py"
+        ),
+        "recovery_development_analyzer": os.path.join(
+            ROOT,
+            "experiments",
+            "analyze_bci_v2_recovery_development.py",
+        ),
+        "recovery_confirmation_analyzer": os.path.join(
+            ROOT,
+            "experiments",
+            "analyze_bci_v2_recovery_confirmation.py",
+        ),
+        "base_dynamics": os.path.join(ROOT, "sdil", "bci.py"),
+        "v2_dynamics": os.path.join(ROOT, "sdil", "bci_v2.py"),
+        "v2_metrics": os.path.join(ROOT, "sdil", "bci_v2_metrics.py"),
+        "recovery_metrics": os.path.join(
+            ROOT, "sdil", "bci_v2_recovery_metrics.py"
+        ),
+        "protocol": PROTOCOL_PATH,
+        "d4_gate": D4_GATE_PATH,
+        "old_r2_gate": OLD_R2_GATE_PATH,
+        "failed_v2_gate": FAILED_V2_GATE_PATH,
+        "failed_target_gate": FAILED_TARGET_GATE_PATH,
+    }
+
+
+def provenance(extra_paths=None):
+    def run(command):
+        return subprocess.run(
+            command,
+            cwd=ROOT,
+            check=True,
+            capture_output=True,
+            text=True,
+        ).stdout.strip()
+
+    paths = source_paths()
+    if extra_paths:
+        paths.update(extra_paths)
+    relative = {
+        name: os.path.relpath(path, ROOT)
+        for name, path in paths.items()
+    }
+    tracked = {
+        name: subprocess.run(
+            ["git", "ls-files", "--error-unmatch", path],
+            cwd=ROOT,
+            capture_output=True,
+        ).returncode == 0
+        for name, path in relative.items()
+    }
+    return {
+        "git_commit": run(["git", "rev-parse", "HEAD"]),
+        "git_tracked_dirty": bool(run([
+            "git", "status", "--porcelain", "--untracked-files=no"
+        ])),
+        "tracked_inputs": tracked,
+        "input_sha256": {
+            name: sha256(path) for name, path in paths.items()
+        },
+    }
+
+
+def require_parent_gates():
+    with open(D4_GATE_PATH) as handle:
+        d4 = json.load(handle)
+    with open(OLD_R2_GATE_PATH) as handle:
+        old_r2 = json.load(handle)
+    with open(FAILED_V2_GATE_PATH) as handle:
+        failed_v2 = json.load(handle)
+    with open(FAILED_TARGET_GATE_PATH) as handle:
+        failed_target = json.load(handle)
+    if not (
+        d4.get("status") == "passed"
+        and d4.get("review_score_after") == 7
+        and old_r2.get("status") == "failed"
+        and failed_v2.get("status") == "failed"
+        and failed_target.get("protocol")
+        == "oral_b_v2_cold_start_recovery_development_v1"
+        and failed_target.get("status") == "failed"
+        and failed_target.get("recovery_confirmation_opened") is False
+        and failed_target.get("review_score_after") == 7
+    ):
+        raise ValueError(
+            "calibrated recovery requires D4 and all preserved failures"
+        )
+
+
+def calibrate_targets(model, task_seed, seed_offset):
+    calibration_seed = seed_offset + task_seed
+    calibration_cfg = replace(model.cfg, target=UNCENSORED_TARGET)
+    trajectories = generate_trajectories(
+        calibration_cfg,
+        calibration_seed,
+        days=1,
+        episodes=CALIBRATION_EPISODES,
+    )
+    calibration_model = model.clone()
+    calibration_model.cfg = calibration_cfg
+    report = run_day_v2(
+        calibration_model,
+        trajectories,
+        0,
+        horizon=calibration_cfg.steps_per_episode,
+        plasticity_gain=0.0,
+        learn_role=False,
+        probe_role=False,
+        learn_predictor=False,
+        learn_critic=False,
+        critic_enabled=False,
+        terminal_outcome_enabled=False,
+        collect=True,
+    )
+    if bool(report["success"].any()):
+        raise RuntimeError("uncensored calibration target was reached")
+    cursors = torch.stack([
+        event["cursor"].double() for event in report["events"]
+    ])
+    maxima = cursors.max(0).values
+    quantiles = torch.tensor(TARGET_QUANTILES, dtype=torch.float64)
+    targets = torch.quantile(maxima, quantiles).tolist()
+    if not all(
+        targets[index] < targets[index + 1]
+        for index in range(len(targets) - 1)
+    ):
+        raise RuntimeError("calibrated targets must be strictly ordered")
+    return tuple(targets), {
+        "seed": calibration_seed,
+        "episodes": CALIBRATION_EPISODES,
+        "quantiles": list(TARGET_QUANTILES),
+        "targets": targets,
+        "uses_outcome_labels": False,
+        "active_state_episode_steps": report["active_transitions"],
+        "maximum_cursor_summary": {
+            "minimum": maxima.min().item(),
+            "median": maxima.median().item(),
+            "maximum": maxima.max().item(),
+        },
+    }
+
+
+def evaluate_calibrated_ladder(
+    model,
+    task_seed,
+    calibration_seed_offset,
+    challenge_seed_offset,
+):
+    targets, calibration = calibrate_targets(
+        model, task_seed, calibration_seed_offset
+    )
+    mode_events = {name: [] for name in ACUTE_MODES}
+    costs = {name: 0 for name in ACUTE_MODES}
+    seeds = {}
+    for target_index, target in enumerate(targets, start=1):
+        challenge_cfg = replace(model.cfg, target=target)
+        trajectory_seed = (
+            challenge_seed_offset
+            + 1_000 * task_seed
+            + target_index - 1
+        )
+        seeds[str(target_index)] = trajectory_seed
+        trajectories = generate_trajectories(
+            challenge_cfg,
+            trajectory_seed,
+            days=1,
+            episodes=CHALLENGE_EPISODES,
+        )
+        for mode in ACUTE_MODES:
+            settings = {
+                "intact": {},
+                "acute_critic_lesion": {"critic_enabled": False},
+                "acute_outcome_lesion": {
+                    "terminal_outcome_enabled": False
+                },
+            }[mode]
+            assay_model = model.clone()
+            assay_model.cfg = challenge_cfg
+            report = run_day_v2(
+                assay_model,
+                trajectories,
+                0,
+                horizon=challenge_cfg.steps_per_episode,
+                plasticity_gain=0.0,
+                learn_role=False,
+                probe_role=False,
+                learn_predictor=False,
+                learn_critic=False,
+                collect=True,
+                **settings,
+            )
+            annotate_target_events(
+                report["events"],
+                0,
+                report["success"],
+                (
+                    3_000_000
+                    + (target_index - 1) * CHALLENGE_EPISODES
+                ),
+                target_index,
+            )
+            mode_events[mode].extend(report["events"])
+            costs[mode] += report["active_transitions"]
+    return targets, mode_events, {
+        "calibration": calibration,
+        "targets": list(targets),
+        "episodes_per_target": CHALLENGE_EPISODES,
+        "trajectory_seeds": seeds,
+        "active_state_episode_steps_by_mode": costs,
+        "selection_over_evaluation_outcomes": False,
+        "maximum_steps_per_episode": model.cfg.steps_per_episode,
+    }
+
+
+def run_cell(
+    task_seed,
+    model_seed,
+    *,
+    split,
+    performance_seed_offset,
+    calibration_seed_offset,
+    challenge_seed_offset,
+):
+    cfg = build_recovery_config()
+    trajectories = generate_trajectories(cfg, task_seed)
+    performance_seed = performance_seed_offset + task_seed
+    performance_trajectories = generate_trajectories(
+        cfg, performance_seed, days=1, episodes=256
+    )
+    initial = BCIV2(cfg, model_seed)
+    conditions = {}
+    trained = {}
+    warmups = {}
+    training_events = None
+    started = time.perf_counter()
+    for name in CONDITIONS:
+        model, warmup = neutral_warmup(
+            initial, task_seed, model_seed, name
+        )
+        events, report = train(
+            model, trajectories, name, collect=name == "intact"
+        )
+        warmups[name] = warmup
+        conditions[name] = report
+        trained[name] = model
+        if name == "intact":
+            training_events = events
+    for name in CONDITIONS:
+        report = evaluate_performance(
+            trained[name], performance_trajectories
+        )
+        conditions[name]["final_success"] = report["success_rate"]
+        conditions[name]["evaluation_active_state_episode_steps"] = (
+            report["active_transitions"]
+        )
+    targets, challenge_events, challenge_cost = (
+        evaluate_calibrated_ladder(
+            trained["intact"],
+            task_seed,
+            calibration_seed_offset,
+            challenge_seed_offset,
+        )
+    )
+    signatures = recovery_signature_metrics(
+        training_events,
+        challenge_events,
+        cfg,
+        trained["intact"].role,
+        targets,
+    )
+    return {
+        "config": vars(cfg),
+        "warmup": warmups,
+        "conditions": conditions,
+        "signatures": signatures,
+        "assays": {
+            "performance_evaluation_seed": performance_seed,
+            "performance_evaluation_episodes": 256,
+            "challenge": challenge_cost,
+        },
+        "wall_s": time.perf_counter() - started,
+        "peak_rss_mib": (
+            resource.getrusage(resource.RUSAGE_SELF).ru_maxrss / 1024
+        ),
+        "hardware": {
+            "device": "cpu",
+            "platform": platform.platform(),
+            "torch_version": torch.__version__,
+            "threads": torch.get_num_threads(),
+        },
+        "split": split,
+    }
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--task-seed", type=int, choices=TASK_SEEDS, required=True
+    )
+    parser.add_argument(
+        "--model-seed", type=int, choices=MODEL_SEEDS, default=0
+    )
+    parser.add_argument(
+        "--outdir", default="results/bci_v2_calibrated_dev"
+    )
+    args = parser.parse_args()
+    require_parent_gates()
+    source = provenance()
+    if (
+        source["git_tracked_dirty"]
+        or not all(source["tracked_inputs"].values())
+    ):
+        raise RuntimeError(
+            "calibrated R1 requires clean, tracked, frozen inputs"
+        )
+    cell = run_cell(
+        args.task_seed,
+        args.model_seed,
+        split="development_validation",
+        performance_seed_offset=540_000,
+        calibration_seed_offset=550_000,
+        challenge_seed_offset=560_000,
+    )
+    result = {
+        "schema_version": 4,
+        "protocol": {
+            "name":
+                "oral_b_v2_calibrated_recovery_development_v1",
+            "selection_split": "development_validation",
+            "hyperparameter_selection": False,
+            "confirmation_seeds_touched": False,
+            "fixed_config": FIXED_CONFIG,
+            "calibration_uses_outcome_labels": False,
+            "protocol_sha256": source["input_sha256"]["protocol"],
+            "failed_target_gate_sha256": source["input_sha256"][
+                "failed_target_gate"
+            ],
+        },
+        "args": vars(args),
+        "provenance": source,
+        **cell,
+    }
+    result["finite"] = finite_tree(result)
+    if not result["finite"]:
+        raise RuntimeError("non-finite calibrated development record")
+    os.makedirs(args.outdir, exist_ok=True)
+    path = os.path.join(
+        args.outdir,
+        f"bci_v2_calibrated_t{args.task_seed}_m0.json",
+    )
+    if os.path.exists(path):
+        raise FileExistsError(f"refusing to overwrite {path}")
+    with open(path, "w") as handle:
+        json.dump(result, handle, indent=2, sort_keys=True)
+        handle.write("\n")
+    print(json.dumps({
+        "path": path,
+        "intact_final": result["conditions"]["intact"]["final_success"],
+        "challenge_success_fraction": result["signatures"][
+            "challenge_success_fraction"
+        ],
+        "terminal_outcome_accuracy": result["signatures"][
+            "terminal_residual_outcome_balanced_acc"
+        ],
+        "finite": result["finite"],
+    }, indent=2))
+
+
+if __name__ == "__main__":
+    main()
diff --git a/experiments/bci_v2_calibrated_smoke.py b/experiments/bci_v2_calibrated_smoke.py
new file mode 100644
index 0000000..1c05783
--- /dev/null
+++ b/experiments/bci_v2_calibrated_smoke.py
@@ -0,0 +1,36 @@
+#!/usr/bin/env python3
+"""Endpoint-free checks for calibration-split target selection."""
+import os
+import sys
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+from experiments.bci_v2_calibrated_run import (
+    CALIBRATION_EPISODES,
+    TARGET_QUANTILES,
+    calibrate_targets,
+)
+from experiments.bci_v2_recovery_run import build_recovery_config
+from sdil.bci_v2 import BCIV2
+
+
+def check_calibration_without_outcomes():
+    cfg = build_recovery_config()
+    model = BCIV2(cfg, model_seed=901)
+    targets, report = calibrate_targets(
+        model, task_seed=902, seed_offset=900_000
+    )
+    assert report["episodes"] == CALIBRATION_EPISODES
+    assert report["quantiles"] == list(TARGET_QUANTILES)
+    assert report["uses_outcome_labels"] is False
+    assert len(targets) == len(TARGET_QUANTILES)
+    assert all(
+        targets[index] < targets[index + 1]
+        for index in range(len(targets) - 1)
+    )
+    assert all(-2.0 < target < 2.0 for target in targets)
+    print("calibration targets use only independent cursor quantiles: exact")
+
+
+if __name__ == "__main__":
+    check_calibration_without_outcomes()
+    print("ALL CALIBRATION-SPLIT MECHANICS CHECKS PASSED")