Add historical exact FOM 100/9 reference

README.md

@@ -17,6 +17,8 @@ configs:
   data_files: data/round_summary.jsonl
 - config_name: stage2_best_verified
   data_files: data/stage2_best_verified.jsonl
+- config_name: historical_exact_fom11
+  data_files: data/historical_exact_fom11.jsonl
 ---
 
 # QLDPC FOM>12 Search Evidence
@@ -37,6 +39,16 @@ The latest source boundary contained 102 raw-score ties; only the 13 released
 rows completed the full replay used here. The dataset does not claim that all
 102 raw ties have the same evidence level.
 
+## Historical exact-distance reference (2026-08-17)
+
+The separate `historical_exact_fom11` configuration contains one
+canonical-deduplicated `[[288,50,8]]` CSS-BB code with legacy exact FOM
+`100/9 ~= 11.1111`. The Campaign 4 source reports `d_X=d_Z=8` and all 100
+logical MILPs optimal. This is a historical exact non-winner, not a current
+Stage 2 result. Its identity and matrices were replayed with the current code,
+but the legacy run did not preserve a portable proof artifact under the modern
+certificate schema. See
+[`docs/HISTORICAL_EXACT_FOM11.md`](docs/HISTORICAL_EXACT_FOM11.md).
 
 ## Snapshot summary
 

data/historical_exact_fom11.jsonl

@@ -0,0 +1 @@
+{"authority":{"modern_replayable_distance_certificate":false,"pipeline_promotion":false,"publication_certificate":false},"candidate_id":"5bf3e9dc496750d28d4e","canonical_digest":"5bf3e9dc496750d28d4e5d70a41fe57494465d03178204e90d9e75367de0200b","code":"[[288,50,8]]","construction":{"A_terms":[[0,0],[1,0],[0,5],[1,5]],"B_terms":[[0,0],[0,1],[5,0],[5,1]],"ell":18,"family":"css-bb","geometry":"rectangular","m":8},"distance":{"d_x":8,"d_z":8,"exact":8,"lower_bound":8,"status":"exact_per_legacy_all_optimal_milp","upper_bound":8},"evidence_level":"legacy_solver_exact_summary_with_modern_identity_replay","fom":{"exact":{"decimal":11.11111111111111,"denominator":9,"numerator":100},"gap_to_12":{"decimal":0.8888888888888888,"denominator":9,"numerator":8}},"identity":{"construction_sha256":"6e518c1b260f96cca3ed61b38df375cba565c086dc3979397d12cc31df33d917","identity_matrix_replay_verified":true,"registry_id":"qcode-css-verified-0660","registry_sha256":"536c733d68d344ed9fef9992bbeb1000900174ad0c500ddb7004126ae5c98408"},"limitations":"The legacy result preserves an all-logicals-optimal MILP summary but not a portable witness/proof artifact under the current hardened certificate schema.","matrices":{"hx_content_sha256":"1cacadd7b31719819d1d5f562e3d38776814fd30bf2bc75fd7f8f39d0ae3900e","hx_file_sha256":"2dc73d9bc5ce64bf98325d343feea4895d68036bcb9cf9882bb71f3633c497a3","hx_path":"matrices/5bf3e9dc496750d28d4e/Hx.mtx.gz","hx_shape":[144,288],"hz_content_sha256":"900df0398e710690b2dac093750ac07dfc04e3461402212292c3469ec94836ee","hz_file_sha256":"70cf6f60310168d951960fadb3c9002539bb7ed98b9f51007eda4de10c9b3ced","hz_path":"matrices/5bf3e9dc496750d28d4e/Hz.mtx.gz","hz_shape":[144,288]},"matrix_checks":{"commutation_violations":0,"css_commutation":true,"k_recomputed":50,"max_check_weight":8,"max_qubit_degree":8,"max_sector_qubit_degree":4,"rank_x":119,"rank_z":119},"parameters":{"k":50,"n":288},"record_kind":"historical_exact_distance_reference","record_sha256":"15b84ad44a3bf9a2e9c97d2be83dab26827780d8e05b753520fc1cb3d3385b38","source_stage":"campaign4_milp_exact","stage":"historical_campaign4","stage_2":false,"status":"LEGACY_MILP_EXACT_NON_WIN","target":{"certified_win":false,"inclusive_required_distance":9,"strict_predicate":"k*d*d > 12*n","strict_required_distance":9}}

docs/HISTORICAL_EXACT_FOM11.md

@@ -0,0 +1,45 @@
+# Historical exact-distance reference with FOM 100/9
+
+This configuration publishes one canonical-deduplicated CSS-BB code:
+
+- parameters: `[[288,50,8]]`;
+- exact FOM in the legacy result: `50*8^2/288 = 100/9 ~= 11.11111111`;
+- source stage: `campaign4_milp_exact`;
+- source status: both sectors reported distance 8, with all 100 logical MILPs optimal;
+- target status: exact non-winner (`100/9 < 12`).
+
+Two stored polynomial orientations reduce to the same canonical CSS code. The
+release therefore contains one row, with provenance for both source rows.
+The supplied `H_X` and `H_Z` matrices were reconstructed again from the
+canonical polynomial representation; their ranks, CSS commutation, and
+canonical registry identity were replayed successfully.
+
+## Evidence boundary
+
+This is a **historical Campaign 4 legacy MILP-exact result**, not a current
+Stage 2 certificate. The legacy JSONL preserves the solver summary
+(`d_x=d_z=8`, 100/100 logical problems optimal), but it does not preserve a
+portable proof/witness artifact bound under the current hardened certificate
+schema. Consequently:
+
+- `exact_distance=8` reports the historical all-optimal MILP result;
+- `modern_replayable_distance_certificate=false`;
+- `publication_certificate=false` and `pipeline_promotion=false`;
+- the row must not be presented as a new Stage 2 strict win.
+
+Run `python -B scripts/verify_historical_exact_fom11.py .` to verify file
+hashes, reconstruct the matrices, check GF(2) ranks and CSS commutation, and
+recompute the rational FOM. That verifier checks the published identity and
+summary bindings; it does not recreate the omitted legacy MILP proof search.
+
+## 中文说明
+
+该配置发布一个规范去重后的 `[[288,50,8]]` CSS-BB 码，历史 MILP 结果中
+精确 FOM 为 `100/9 ~= 11.1111`，因此它是 exact non-winner。两个 X/Z
+方向共 100 个逻辑问题均记录为 optimal。矩阵、秩、对易关系和 registry
+身份已经重新回放核验。
+
+但这条记录来自历史 Campaign 4，并非当前 Stage 2 的现代证书；旧结果未
+保存符合当前 hardened schema 的可移植 proof artifact。因此数据中明确
+保留 `modern_replayable_distance_certificate=false`，不能把它描述成新的
+Stage 2 认证胜者。

evidence/historical_exact_fom11.jsonl

@@ -0,0 +1 @@
+{"authority":{"modern_replayable_distance_certificate":false,"pipeline_promotion":false,"publication_certificate":false},"candidate_id":"5bf3e9dc496750d28d4e","canonical_digest":"5bf3e9dc496750d28d4e5d70a41fe57494465d03178204e90d9e75367de0200b","evidence_sha256":"932d161aaa515fe049bec9f6c37d5c1d240347ba5c2ae1ee6741d6e209a0c536","identity_replay":{"canonical_digest":"5bf3e9dc496750d28d4e5d70a41fe57494465d03178204e90d9e75367de0200b","construction_sha256":"6e518c1b260f96cca3ed61b38df375cba565c086dc3979397d12cc31df33d917","css_commutation":true,"k_recomputed":50,"registry_id":"qcode-css-verified-0660","verified":true},"legacy_milp_summary":{"d_x":8,"d_x_computed":true,"d_z":8,"exact":true,"logicals_incumbent":0,"logicals_optimal":100,"num_logicals_checked":100,"timeout_per_logical_s":300,"total_logicals":100},"limitations":{"distance_claim":"Exact under the legacy Campaign 4 per-logical MILP result semantics.","portable_modern_certificate":"Not available: the source JSONL did not persist a modern bound-instance proof, solver-runtime binding, or independently replayable distance certificate.","safe_use":"Historical exact-distance reference and identity/matrix reproduction; not a current Stage 2 promotion artifact."},"record_kind":"historical_exact_distance_evidence","source":{"file_sha256":"763f0e57f313b6d54dd2232a2c7e5bb3de08bb5f1e8757844715fc41b90408d9","path":"results/campaign4_milp_verified.jsonl","records":[{"elapsed_s":4384.728747079003,"orientation":"canonical","raw_line_sha256":"9b3f9697c6c56f6c46dcbcba358b57f6f806306b91925f79ceaa797c7c642fd5","row_index_zero_based":660,"saved_at":"2026-03-09T10:36:27+00:00"},{"elapsed_s":4883.398503006996,"orientation":"equivalent_A_B_swapped_form","raw_line_sha256":"3eb80f4c23f371977736fe0714266dd43ad610593afc87a158a802e31321d8ab","row_index_zero_based":676,"saved_at":"2026-03-09T10:44:46+00:00"}]},"source_stage":"campaign4_milp_exact","stage_2":false}

manifests/SHA256SUMS

@@ -1,6 +1,6 @@
 ffe79769ee48619a33c8049eef31be7f633ca00cf0fa1d4d852cbd1e4f9d8936  CITATION.cff
 c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4  LICENSE
-21f54159c99a91ea797a65d0b1882f93bd4c1eeba6dce556e39eff2cd431ffcb  README.md
+48cb87c52461a7e60cab860750bbfff717f4aaaea3892a3725e12480e72f748c  README.md
 95a5ec2a8c23f4aab17998c6fad05490a316c083983d7183feaad075e3dff397  THIRD_PARTY_NOTICES.md
 9476a8f02cd6c3e09e0ff73f4d5013f3b98902b3d157072c233f5012cd60d00b  config/run_config.public.json
 6db5c6d622831117e5e8276d9d87d22b01889082b1f6fd74b770ed1b4abb6edb  data/candidates.jsonl
@@ -122,3 +122,10 @@ a910a0b82c8e86b8784052f038e8287f3a45f2f6e807cd9804d293cddd07a385  data/stage2_be
 e20a073c69c2fa9fdb8182063ab592c3f3c757323b2d116142b3befc61d43034  docs/STAGE2_BEST_RESULT.md
 0aea4cf3413f06796db3b0c5f4bae97712c35db6b834f419a59e77bd10e00f9f  evidence/stage2_best_verified.jsonl
 21669a77c80405f611407140fbe5b9b6895db729e1deae61eb93d23675f7178c  manifests/stage2_best_verified.json
+630a79802d339c69ce3008be586d2e3c11fc1baefc43aa49282fe2b73eda6cf3  data/historical_exact_fom11.jsonl
+a1d1a21f66b3c53858627dbe76b87bfe96756d0eb20471a913e71ecce76e4917  docs/HISTORICAL_EXACT_FOM11.md
+77e73fe488d9204044aa9e64af655db9f2f35eb0aefaf4fcc108d2b7f31c19e1  evidence/historical_exact_fom11.jsonl
+a213484ae5bf13db227919e3edbfdbac3cb47745afa58f9abe31ee5cd77e3b20  manifests/historical_exact_fom11.json
+2dc73d9bc5ce64bf98325d343feea4895d68036bcb9cf9882bb71f3633c497a3  matrices/5bf3e9dc496750d28d4e/Hx.mtx.gz
+70cf6f60310168d951960fadb3c9002539bb7ed98b9f51007eda4de10c9b3ced  matrices/5bf3e9dc496750d28d4e/Hz.mtx.gz
+1c5c46b3ee1893effa7b01009f8681e881e0e9e13f6a7f17ced21adba2d4e12d  scripts/verify_historical_exact_fom11.py

manifests/historical_exact_fom11.json

@@ -0,0 +1,71 @@
+{
+  "authority": {
+    "modern_replayable_distance_certificate": false,
+    "pipeline_promotion": false,
+    "publication_certificate": false
+  },
+  "candidate": {
+    "canonical_digest": "5bf3e9dc496750d28d4e5d70a41fe57494465d03178204e90d9e75367de0200b",
+    "code": "[[288,50,8]]",
+    "exact_fom_denominator": 9,
+    "exact_fom_numerator": 100,
+    "registry_id": "qcode-css-verified-0660"
+  },
+  "created_date_utc": "2026-08-17",
+  "dataset_base_commit": "609c04215f4a656bf477a5c255596a2fd6ea4218",
+  "files": [
+    {
+      "bytes": 2291,
+      "path": "data/historical_exact_fom11.jsonl",
+      "rows": 1,
+      "sha256": "630a79802d339c69ce3008be586d2e3c11fc1baefc43aa49282fe2b73eda6cf3"
+    },
+    {
+      "bytes": 1924,
+      "path": "evidence/historical_exact_fom11.jsonl",
+      "rows": 1,
+      "sha256": "77e73fe488d9204044aa9e64af655db9f2f35eb0aefaf4fcc108d2b7f31c19e1"
+    },
+    {
+      "bytes": 2928,
+      "path": "matrices/5bf3e9dc496750d28d4e/Hx.mtx.gz",
+      "sha256": "2dc73d9bc5ce64bf98325d343feea4895d68036bcb9cf9882bb71f3633c497a3"
+    },
+    {
+      "bytes": 2812,
+      "path": "matrices/5bf3e9dc496750d28d4e/Hz.mtx.gz",
+      "sha256": "70cf6f60310168d951960fadb3c9002539bb7ed98b9f51007eda4de10c9b3ced"
+    },
+    {
+      "bytes": 2249,
+      "path": "docs/HISTORICAL_EXACT_FOM11.md",
+      "sha256": "a1d1a21f66b3c53858627dbe76b87bfe96756d0eb20471a913e71ecce76e4917"
+    },
+    {
+      "bytes": 6947,
+      "path": "scripts/verify_historical_exact_fom11.py",
+      "sha256": "1c5c46b3ee1893effa7b01009f8681e881e0e9e13f6a7f17ced21adba2d4e12d"
+    }
+  ],
+  "gate": "qcode-historical-campaign4-exact-reference-v1",
+  "manifest_sha256": "e4e883902830ba1339de52e6630b4c91d5c97e896e9310cb807de213477445c5",
+  "record_kind": "historical_exact_distance_reference",
+  "schema_version": 1,
+  "source": {
+    "file_sha256": "763f0e57f313b6d54dd2232a2c7e5bb3de08bb5f1e8757844715fc41b90408d9",
+    "path": "results/campaign4_milp_verified.jsonl",
+    "records": [
+      {
+        "raw_line_sha256": "9b3f9697c6c56f6c46dcbcba358b57f6f806306b91925f79ceaa797c7c642fd5",
+        "row_index_zero_based": 660
+      },
+      {
+        "raw_line_sha256": "3eb80f4c23f371977736fe0714266dd43ad610593afc87a158a802e31321d8ab",
+        "row_index_zero_based": 676
+      }
+    ],
+    "source_stage": "campaign4_milp_exact"
+  },
+  "stage_2": false,
+  "status": "LEGACY_MILP_EXACT_NON_WIN"
+}

matrices/5bf3e9dc496750d28d4e/Hx.mtx.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2dc73d9bc5ce64bf98325d343feea4895d68036bcb9cf9882bb71f3633c497a3
+size 2928

matrices/5bf3e9dc496750d28d4e/Hz.mtx.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70cf6f60310168d951960fadb3c9002539bb7ed98b9f51007eda4de10c9b3ced
+size 2812

scripts/verify_historical_exact_fom11.py

@@ -0,0 +1,192 @@
+#!/usr/bin/env python3
+from __future__ import annotations
+
+import gzip
+import hashlib
+import json
+import sys
+from pathlib import Path
+
+
+def sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as handle:
+        for block in iter(lambda: handle.read(1 << 20), b""):
+            digest.update(block)
+    return digest.hexdigest()
+
+
+def canonical_sha256(value: dict, omitted: str) -> str:
+    payload = {key: item for key, item in value.items() if key != omitted}
+    encoded = json.dumps(payload, sort_keys=True, separators=(",", ":"))
+    return hashlib.sha256(encoded.encode()).hexdigest()
+
+
+def load_jsonl_one(path: Path) -> dict:
+    lines = [line for line in path.read_text().splitlines() if line.strip()]
+    assert len(lines) == 1, (path, len(lines))
+    return json.loads(lines[0])
+
+
+def load_mtx(path: Path) -> list[list[int]]:
+    with gzip.open(path, "rt", encoding="ascii") as handle:
+        assert handle.readline().strip() == (
+            "%%MatrixMarket matrix coordinate integer general"
+        )
+        line = handle.readline()
+        while line.startswith("%"):
+            line = handle.readline()
+        rows, cols, nnz = map(int, line.split())
+        matrix = [[0] * cols for _ in range(rows)]
+        seen = 0
+        for line in handle:
+            if not line.strip():
+                continue
+            row, col, value = map(int, line.split())
+            assert value == 1
+            matrix[row - 1][col - 1] ^= 1
+            seen += 1
+        assert seen == nnz
+        return matrix
+
+
+def gf2_rank(matrix: list[list[int]]) -> int:
+    packed = [
+        sum((value & 1) << col for col, value in enumerate(row))
+        for row in matrix
+    ]
+    rank = 0
+    while packed:
+        pivot = max(packed)
+        if not pivot:
+            break
+        bit = 1 << (pivot.bit_length() - 1)
+        rank += 1
+        packed = [
+            value ^ pivot if value & bit else value
+            for value in packed
+            if value != pivot
+        ]
+    return rank
+
+
+def content_sha256(matrix: list[list[int]]) -> str:
+    shape = [len(matrix), len(matrix[0]) if matrix else 0]
+    header = json.dumps(
+        {"dtype": "uint8-gf2", "shape": shape},
+        sort_keys=True,
+        separators=(",", ":"),
+    )
+    raw = bytes(value & 1 for row in matrix for value in row)
+    return hashlib.sha256(header.encode() + b"\n" + raw).hexdigest()
+
+
+def polynomial_matrix(
+    ell: int,
+    m: int,
+    terms: list[list[int]],
+) -> list[list[int]]:
+    size = ell * m
+    matrix = [[0] * size for _ in range(size)]
+    for row in range(size):
+        x, y = divmod(row, m)
+        for dx, dy in terms:
+            col = ((x + dx) % ell) * m + ((y + dy) % m)
+            matrix[row][col] ^= 1
+    return matrix
+
+
+def transpose(matrix: list[list[int]]) -> list[list[int]]:
+    return [list(column) for column in zip(*matrix)]
+
+
+def hstack(
+    left: list[list[int]],
+    right: list[list[int]],
+) -> list[list[int]]:
+    return [a + b for a, b in zip(left, right)]
+
+
+def parity(left: list[int], right: list[int]) -> int:
+    return sum(a & b for a, b in zip(left, right)) & 1
+
+
+def main(root: Path) -> None:
+    root = root.resolve()
+    data_path = root / "data/historical_exact_fom11.jsonl"
+    evidence_path = root / "evidence/historical_exact_fom11.jsonl"
+    manifest_path = root / "manifests/historical_exact_fom11.json"
+    row = load_jsonl_one(data_path)
+    evidence = load_jsonl_one(evidence_path)
+    manifest = json.loads(manifest_path.read_text())
+
+    assert canonical_sha256(row, "record_sha256") == row["record_sha256"]
+    assert canonical_sha256(evidence, "evidence_sha256") == evidence["evidence_sha256"]
+    assert canonical_sha256(manifest, "manifest_sha256") == manifest["manifest_sha256"]
+    for item in manifest["files"]:
+        path = root / item["path"]
+        assert path.stat().st_size == item["bytes"]
+        assert sha256(path) == item["sha256"]
+
+    assert row["canonical_digest"] == evidence["canonical_digest"]
+    construction = row["construction"]
+    construction_payload = {
+        key: construction[key]
+        for key in ("ell", "m", "A_terms", "B_terms")
+    }
+    encoded = json.dumps(
+        construction_payload,
+        sort_keys=True,
+        separators=(",", ":"),
+    )
+    assert hashlib.sha256(encoded.encode()).hexdigest() == row["identity"]["construction_sha256"]
+
+    ell, m = construction["ell"], construction["m"]
+    a_matrix = polynomial_matrix(ell, m, construction["A_terms"])
+    b_matrix = polynomial_matrix(ell, m, construction["B_terms"])
+    expected_hx = hstack(a_matrix, b_matrix)
+    expected_hz = hstack(transpose(b_matrix), transpose(a_matrix))
+    matrices = row["matrices"]
+    hx_path = root / matrices["hx_path"]
+    hz_path = root / matrices["hz_path"]
+    hx, hz = load_mtx(hx_path), load_mtx(hz_path)
+    assert hx == expected_hx and hz == expected_hz
+    assert [len(hx), len(hx[0])] == matrices["hx_shape"] == [144, 288]
+    assert [len(hz), len(hz[0])] == matrices["hz_shape"] == [144, 288]
+    assert sha256(hx_path) == matrices["hx_file_sha256"]
+    assert sha256(hz_path) == matrices["hz_file_sha256"]
+    assert content_sha256(hx) == matrices["hx_content_sha256"]
+    assert content_sha256(hz) == matrices["hz_content_sha256"]
+    rank_x, rank_z = gf2_rank(hx), gf2_rank(hz)
+    assert rank_x == row["matrix_checks"]["rank_x"] == 119
+    assert rank_z == row["matrix_checks"]["rank_z"] == 119
+    n = row["parameters"]["n"]
+    k = row["parameters"]["k"]
+    distance = row["distance"]["exact"]
+    max_check_weight = max(max(map(sum, hx)), max(map(sum, hz)))
+    max_sector_qubit_degree = max(
+        max(sum(check[j] for check in hx) for j in range(n)),
+        max(sum(check[j] for check in hz) for j in range(n)),
+    )
+    max_qubit_degree = max(
+        sum(check[j] for check in hx) + sum(check[j] for check in hz)
+        for j in range(n)
+    )
+    assert row["matrix_checks"]["max_check_weight"] == max_check_weight == 8
+    assert row["matrix_checks"]["max_sector_qubit_degree"] == max_sector_qubit_degree == 4
+    assert row["matrix_checks"]["max_qubit_degree"] == max_qubit_degree == 8
+    assert n - rank_x - rank_z == k == 50
+    assert all(parity(x_row, z_row) == 0 for x_row in hx for z_row in hz)
+    assert row["fom"]["exact"]["numerator"] == 100
+    assert row["fom"]["exact"]["denominator"] == 9
+    assert k * distance * distance * 9 == 100 * n
+    assert k * distance * distance < 12 * n
+    assert row["authority"]["modern_replayable_distance_certificate"] is False
+    assert evidence["legacy_milp_summary"]["logicals_optimal"] == 100
+    assert evidence["legacy_milp_summary"]["total_logicals"] == 100
+    assert evidence["source"]["records"][1]["orientation"] == "equivalent_A_B_swapped_form"
+    print("PASS: historical [[288,50,8]] identity, matrices, hashes, and FOM verified")
+
+
+if __name__ == "__main__":
+    main(Path(sys.argv[1]) if len(sys.argv) > 1 else Path("."))