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

1 files changed, 217 insertions, 0 deletions

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()