test(performance): enforce semantic artifact authority boundary

MODULE_SPEC.yaml

@@ -1,7 +1,7 @@
 {
   "schema_version": 1,
   "module_id": "quant_engine",
-  "authority": {"scope": "module_metadata", "subject": "quant_engine", "owner": "quant-engine-owner", "source": "MODULE_SPEC.yaml", "revision": 4, "effective_from": "2026-09-01T00:00:00+08:00"},
+  "authority": {"scope": "module_metadata", "subject": "quant_engine", "owner": "quant-engine-owner", "source": "MODULE_SPEC.yaml", "revision": 5, "effective_from": "2026-09-01T00:00:00+08:00"},
   "repository": {"name": "quant_engine", "workspace_id": "researchhub", "type": "research_engine", "maturity": "operational"},
   "bounded_context": {
     "domain": "quantitative-research-engine",
@@ -19,7 +19,7 @@
     {"id": "factor-and-indicator-calculation", "summary": "Calculate reusable alpha factors and technical indicators from caller-supplied data.", "status": "operational"},
     {"id": "execution-simulation", "summary": "Simulate costs, slippage, market constraints, fills, NAV, and PnL without live order routing.", "status": "operational"},
     {"id": "portfolio-backtesting", "summary": "Run weight-based backtests and benchmark comparisons.", "status": "operational"},
-    {"id": "backtest-evidence-contracts", "summary": "Identify governed offline backtest inputs and close existing research artifact evidence without persistence or decision authority.", "status": "operational"},
+    {"id": "backtest-evidence-contracts", "summary": "Identify governed offline backtest inputs and close existing research artifact and performance-methodology evidence without recomputation, persistence, or decision authority.", "status": "operational"},
     {"id": "portfolio-risk-computation-contracts", "summary": "Verify deterministic portfolio-computation receipts and expose S3-bound portfolio decisions and risk assessments without adding algorithms or execution authority.", "status": "operational"},
     {"id": "risk-and-performance-analysis", "summary": "Calculate portfolio decomposition, risk contribution, and performance statistics.", "status": "operational"}
   ],
@@ -33,6 +33,7 @@
       {"contract_id": "researchhub.factor-set-ref", "version": "1.0.0", "authority": "quant_engine", "path": "src/quant_engine/factor_contracts.py"},
       {"contract_id": "researchhub.backtest-run-ref", "version": "1.0.0", "authority": "quant_engine", "path": "src/quant_engine/governed_pipeline.py"},
       {"contract_id": "researchhub.backtest-evidence-manifest", "version": "1.0.0", "authority": "quant_engine", "path": "src/quant_engine/artifact.py"},
+      {"contract_id": "researchhub.performance-evidence", "version": "1.0.0", "authority": "quant_engine", "path": "src/quant_engine/artifact.py"},
       {"contract_id": "researchhub.portfolio-decision", "version": "1.0.0", "authority": "quant_engine", "path": "src/quant_engine/portfolio_risk_contracts.py"},
       {"contract_id": "researchhub.risk-assessment", "version": "1.0.0", "authority": "quant_engine", "path": "src/quant_engine/portfolio_risk_contracts.py"}
     ],

README.md

@@ -260,6 +260,35 @@ digest 等价，也不会把旧 run 静默升级为新合同。
 `LEGACY_EXPLORATORY`；不能隐式提升为 `CONTRACT_QUALIFIED`。所有资格均只描述离线证据闭合，
 不表示投资有效、组合获批、Paper、生产或实盘就绪。
 
+## 绩效证据与方法论合同 v1
+
+`quant_engine.artifact.PerformanceEvidenceV1` 在现有计算和事实表之外增加一层只读、内容寻址的
+owner 证据。`build_performance_evidence()` 只接受同一运行的完整 `ResearchRunArtifact`、
+`BacktestRunRef` 与 `CONTRACT_QUALIFIED BacktestEvidenceManifest`；它核对全部 artifact 表、
+performance 表和唯一行摘要，并绑定 artifact、row 与严格对齐 benchmark series 的独立摘要。
+生产 builder 不重算、填补、重命名或覆盖任何绩效值。
+
+方法论固定为日简单收益、252 期年化、绝对指标年化无风险利率 `0.0`、benchmark 日无风险利率
+`0.0`，以及 benchmark 存在时的 `exact_session_index`。相对指标使用封闭 availability：无基准为
+`benchmark_absent`；active variance、benchmark variance 或 alpha 几何年化域不足时分别使用
+对应 `not_estimable_*` 原因。benchmark 存在时 tracking error 始终必须是有限非负值；null 不会
+被转成零。
+
+```python
+from quant_engine.artifact import build_performance_evidence
+
+performance_evidence = build_performance_evidence(
+    artifact,
+    backtest_run_ref,
+    backtest_evidence_manifest,
+)
+canonical_bytes = performance_evidence.canonical_bytes()
+```
+
+该合同范围固定为 `offline_research_only`。它不授予排名、推荐、决策、发布、论文、Paper、生产、
+实盘、交易或投资建议权限，也不包含原始参数、returns、NAV、benchmark series、表字节、存储
+locator、URI 或凭证。
+
 ## 组合决策与风险评估合同 v1
 
 `quant_engine.portfolio_risk_contracts` 是现有计算 owner 外围的薄合同层。创建

tests/fixtures/performance-evidence-v1.golden.json

@@ -0,0 +1,871 @@
+{
+  "cases": {
+    "absent": {
+      "artifact_available_at": "2026-01-08T02:05:00Z",
+      "authority": "quant_engine",
+      "backtest_evidence_manifest_document_sha256": "sha256:b7e9139e021de3122bee9376a8c8387fca3db75fe2a698adccf33471caf971d2",
+      "backtest_evidence_manifest_evidence_digest": "sha256:fae26a93d754e98f437bf9e4b635fc1cdc4f85d004a4bde396823b52e2115de5",
+      "backtest_evidence_manifest_id": "rhbacktestevidencev1:sha256:3f67fcad23c75684ffeb325d8405b2f81139a732e237d30fbb30d23f91e32726",
+      "backtest_evidence_qualification": "contract_qualified",
+      "backtest_run_ref_document_sha256": "sha256:6a798adb3e0568aca84ed3e3a285b92d181ec569462fc003952a8c0d8382ae3a",
+      "backtest_run_ref_id": "rhbacktestrunv1:sha256:5036c771c44a2adade9ea590eee0d8cf824ff8f519fadd8ba424e5d914856386",
+      "benchmark_alignment_policy": "none",
+      "benchmark_id": "",
+      "benchmark_series_digest": null,
+      "calendar": "CN-A",
+      "code_revision": "dddddddddddddddddddddddddddddddddddddddd",
+      "configuration_digest": "sha256:d28b49ea1bebc78d0023667f4fb9990b1fb45807176c2193b458525def056f4d",
+      "cost_model_digest": "sha256:8888888888888888888888888888888888888888888888888888888888888888",
+      "cost_model_version": "1.0.0",
+      "dataset_content_digest": "sha256:44ea11ba64dc2e6fd55c6d8e038c5edc84ee38d6fd46e38b15a1d5409662a020",
+      "dataset_manifest_digest": "sha256:d991bb2f8f6b80525f93c51e0b371213a3ed4649dffb073ed4605bfbd32349bd",
+      "dataset_snapshot_id": "rhdsv1:sha256:f63a29b4795c63fb7d6b2d3b5544cee9274b633db77c75c63d50a340c0827d57",
+      "document_sha256": "sha256:729852af307fd4a94ac566ae45df2e57b3d45c4fbdf45820351b5ef19cdf0ad3",
+      "end_date": "2026-01-08",
+      "environment_lock_digest": "sha256:9999999999999999999999999999999999999999999999999999999999999999",
+      "execution_model_digest": "sha256:7777777777777777777777777777777777777777777777777777777777777777",
+      "execution_model_version": "1.0.0",
+      "factor_output_content_digest": "sha256:d78751460dc27fc796163c462d946924ce2bcb45dcfceb5e4b77f76093befc9e",
+      "factor_set_digest": "sha256:e9339581cf569e92459f672e8081337712e7bf98e58ad42d60d7ed13f9b5a021",
+      "factor_set_id": "rhfactorsetv1:sha256:e9339581cf569e92459f672e8081337712e7bf98e58ad42d60d7ed13f9b5a021",
+      "foundation_digest": "sha256:d848237ab753ee9432ae78ec1f93b6ac45c8072d6694023b7f288203daf9d838",
+      "foundation_id": "rhdfv1:sha256:d848237ab753ee9432ae78ec1f93b6ac45c8072d6694023b7f288203daf9d838",
+      "frequency": "1d",
+      "methodology": {
+        "alpha": "daily_ols_intercept_geometric_annualization",
+        "annual_risk_free": 0.0,
+        "annualized_return": "geometric_compound",
+        "annualized_volatility": "sample_std_sqrt_periods",
+        "benchmark_alignment": "none",
+        "benchmark_risk_free_daily": 0.0,
+        "beta": "sample_covariance_over_sample_variance",
+        "calmar_ratio": "unadjusted_annualized_return_over_absolute_maximum_drawdown",
+        "code_revision": "dddddddddddddddddddddddddddddddddddddddd",
+        "implementation_module": "quant_engine.metrics",
+        "implementation_version": "researchhub.quant-performance-methodology.v1",
+        "information_ratio": "mean_active_over_sample_std_active_sqrt_periods",
+        "maximum_drawdown": "non_positive_peak_to_trough_ratio_with_initial_nav_one",
+        "methodology_id": "researchhub.quant-performance-methodology.v1",
+        "periods_per_year": 252,
+        "return_type": "simple",
+        "sharpe_ratio": "annualized_return_minus_annual_risk_free_over_annualized_volatility",
+        "sortino_ratio": "annualized_return_minus_annual_risk_free_over_root_mean_square_negative_returns_sqrt_periods",
+        "source_frequency": "1d",
+        "total_return": "final_nav_minus_one",
+        "tracking_error": "sample_std_active_return_sqrt_periods",
+        "win_rate": "positive_daily_return_count_over_observation_count"
+      },
+      "methodology_id": "researchhub.quant-performance-methodology.v1",
+      "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+      "metrics": [
+        {
+          "availability": "available",
+          "key": "total_return",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "total_ret",
+          "unit": "ratio",
+          "value": 0.575
+        },
+        {
+          "availability": "available",
+          "key": "annualized_return",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "ann_ret",
+          "unit": "ratio_per_year",
+          "value": 2683336646708.1
+        },
+        {
+          "availability": "available",
+          "key": "annualized_volatility",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "ann_volatility",
+          "unit": "ratio_per_year",
+          "value": 1.38901943830891
+        },
+        {
+          "availability": "available",
+          "key": "sharpe_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "sharpe",
+          "unit": "ratio",
+          "value": 1931820803008.3313
+        },
+        {
+          "availability": "available",
+          "key": "sortino_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "sortino",
+          "unit": "ratio",
+          "value": 0.0
+        },
+        {
+          "availability": "available",
+          "key": "maximum_drawdown",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "max_dd",
+          "unit": "ratio",
+          "value": 0.0
+        },
+        {
+          "availability": "available",
+          "key": "calmar_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "calmar",
+          "unit": "ratio",
+          "value": 0.0
+        },
+        {
+          "availability": "available",
+          "key": "win_rate",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "win_rate",
+          "unit": "ratio",
+          "value": 0.75
+        },
+        {
+          "availability": "benchmark_absent",
+          "key": "tracking_error",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "tracking_error",
+          "unit": "ratio_per_year",
+          "value": null
+        },
+        {
+          "availability": "benchmark_absent",
+          "key": "information_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "ir",
+          "unit": "ratio",
+          "value": null
+        },
+        {
+          "availability": "benchmark_absent",
+          "key": "alpha",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "alpha",
+          "unit": "ratio_per_year",
+          "value": null
+        },
+        {
+          "availability": "benchmark_absent",
+          "key": "beta",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "beta",
+          "unit": "ratio",
+          "value": null
+        },
+        {
+          "availability": "available",
+          "key": "trade_count",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "n_trades",
+          "unit": "count",
+          "value": 3
+        },
+        {
+          "availability": "available",
+          "key": "day_count",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "n_days",
+          "unit": "count",
+          "value": 4
+        }
+      ],
+      "performance_evidence_id": "rhperformanceevidencev1:sha256:d735abe76c28db429e107cb5ba6a1929c2c0f2084a6c4ac41b38ffbc02b2d2bd",
+      "performance_row_digest": "sha256:4e6329b0db4731513fe793491a163dd358e3a6f7baa84d3c94a4d4d2355c7a7d",
+      "performance_table_content_digest": "sha256:074b8511ff9a2154235e4dde8bac300658cb69f900e49f332d61560b82be9af5",
+      "performance_table_logical_name": "performance",
+      "performance_table_row_count": 1,
+      "performance_table_schema_digest": "sha256:16cef93a679761103ae405e622b7929abbfe07115bf276be4f164da5a16128d0",
+      "research_artifact_content_digest": "sha256:ad85ee1317bd8ce5fbef8fddc268b91be8678701ba5b8fbf6d61bf670f918bf9",
+      "research_artifact_schema_version": "1.1.0",
+      "run_id": "rhbacktestrunv1:sha256:5036c771c44a2adade9ea590eee0d8cf824ff8f519fadd8ba424e5d914856386",
+      "schema_version": "researchhub.performance-evidence.v1",
+      "scope": "offline_research_only",
+      "start_date": "2026-01-05",
+      "strategy_digest": "sha256:6666666666666666666666666666666666666666666666666666666666666666",
+      "strategy_id": "alpha-top1",
+      "strategy_version": "1.0.0",
+      "timezone": "Asia/Shanghai"
+    },
+    "estimable": {
+      "artifact_available_at": "2026-01-08T02:05:00Z",
+      "authority": "quant_engine",
+      "backtest_evidence_manifest_document_sha256": "sha256:e2e0e63fb4c733d2dba9a290511b6c8b0132bfe877dfe89ee7e614b74b87bc51",
+      "backtest_evidence_manifest_evidence_digest": "sha256:f3913894d032c389c64eef59058b3cbc694cb9cc14ce9cee2699f0068200b650",
+      "backtest_evidence_manifest_id": "rhbacktestevidencev1:sha256:f94733e849433f62f3da1e1ec8999891b93d49c7d657832d64e64bef9e211117",
+      "backtest_evidence_qualification": "contract_qualified",
+      "backtest_run_ref_document_sha256": "sha256:6a798adb3e0568aca84ed3e3a285b92d181ec569462fc003952a8c0d8382ae3a",
+      "backtest_run_ref_id": "rhbacktestrunv1:sha256:5036c771c44a2adade9ea590eee0d8cf824ff8f519fadd8ba424e5d914856386",
+      "benchmark_alignment_policy": "exact_session_index",
+      "benchmark_id": "000300.SH",
+      "benchmark_series_digest": "sha256:1c564e4d12af0d59b1fd707d7816da2895238969d9207470984db63cfeac423a",
+      "calendar": "CN-A",
+      "code_revision": "dddddddddddddddddddddddddddddddddddddddd",
+      "configuration_digest": "sha256:d28b49ea1bebc78d0023667f4fb9990b1fb45807176c2193b458525def056f4d",
+      "cost_model_digest": "sha256:8888888888888888888888888888888888888888888888888888888888888888",
+      "cost_model_version": "1.0.0",
+      "dataset_content_digest": "sha256:44ea11ba64dc2e6fd55c6d8e038c5edc84ee38d6fd46e38b15a1d5409662a020",
+      "dataset_manifest_digest": "sha256:d991bb2f8f6b80525f93c51e0b371213a3ed4649dffb073ed4605bfbd32349bd",
+      "dataset_snapshot_id": "rhdsv1:sha256:f63a29b4795c63fb7d6b2d3b5544cee9274b633db77c75c63d50a340c0827d57",
+      "document_sha256": "sha256:a6a1a3313f511505c0cf57fddeebdf20a3b0f83a2958f31332d5597d172165e7",
+      "end_date": "2026-01-08",
+      "environment_lock_digest": "sha256:9999999999999999999999999999999999999999999999999999999999999999",
+      "execution_model_digest": "sha256:7777777777777777777777777777777777777777777777777777777777777777",
+      "execution_model_version": "1.0.0",
+      "factor_output_content_digest": "sha256:d78751460dc27fc796163c462d946924ce2bcb45dcfceb5e4b77f76093befc9e",
+      "factor_set_digest": "sha256:e9339581cf569e92459f672e8081337712e7bf98e58ad42d60d7ed13f9b5a021",
+      "factor_set_id": "rhfactorsetv1:sha256:e9339581cf569e92459f672e8081337712e7bf98e58ad42d60d7ed13f9b5a021",
+      "foundation_digest": "sha256:d848237ab753ee9432ae78ec1f93b6ac45c8072d6694023b7f288203daf9d838",
+      "foundation_id": "rhdfv1:sha256:d848237ab753ee9432ae78ec1f93b6ac45c8072d6694023b7f288203daf9d838",
+      "frequency": "1d",
+      "methodology": {
+        "alpha": "daily_ols_intercept_geometric_annualization",
+        "annual_risk_free": 0.0,
+        "annualized_return": "geometric_compound",
+        "annualized_volatility": "sample_std_sqrt_periods",
+        "benchmark_alignment": "exact_session_index",
+        "benchmark_risk_free_daily": 0.0,
+        "beta": "sample_covariance_over_sample_variance",
+        "calmar_ratio": "unadjusted_annualized_return_over_absolute_maximum_drawdown",
+        "code_revision": "dddddddddddddddddddddddddddddddddddddddd",
+        "implementation_module": "quant_engine.metrics",
+        "implementation_version": "researchhub.quant-performance-methodology.v1",
+        "information_ratio": "mean_active_over_sample_std_active_sqrt_periods",
+        "maximum_drawdown": "non_positive_peak_to_trough_ratio_with_initial_nav_one",
+        "methodology_id": "researchhub.quant-performance-methodology.v1",
+        "periods_per_year": 252,
+        "return_type": "simple",
+        "sharpe_ratio": "annualized_return_minus_annual_risk_free_over_annualized_volatility",
+        "sortino_ratio": "annualized_return_minus_annual_risk_free_over_root_mean_square_negative_returns_sqrt_periods",
+        "source_frequency": "1d",
+        "total_return": "final_nav_minus_one",
+        "tracking_error": "sample_std_active_return_sqrt_periods",
+        "win_rate": "positive_daily_return_count_over_observation_count"
+      },
+      "methodology_id": "researchhub.quant-performance-methodology.v1",
+      "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+      "metrics": [
+        {
+          "availability": "available",
+          "key": "total_return",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "total_ret",
+          "unit": "ratio",
+          "value": 0.575
+        },
+        {
+          "availability": "available",
+          "key": "annualized_return",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "ann_ret",
+          "unit": "ratio_per_year",
+          "value": 2683336646708.1
+        },
+        {
+          "availability": "available",
+          "key": "annualized_volatility",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "ann_volatility",
+          "unit": "ratio_per_year",
+          "value": 1.38901943830891
+        },
+        {
+          "availability": "available",
+          "key": "sharpe_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "sharpe",
+          "unit": "ratio",
+          "value": 1931820803008.3313
+        },
+        {
+          "availability": "available",
+          "key": "sortino_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "sortino",
+          "unit": "ratio",
+          "value": 0.0
+        },
+        {
+          "availability": "available",
+          "key": "maximum_drawdown",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "max_dd",
+          "unit": "ratio",
+          "value": 0.0
+        },
+        {
+          "availability": "available",
+          "key": "calmar_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "calmar",
+          "unit": "ratio",
+          "value": 0.0
+        },
+        {
+          "availability": "available",
+          "key": "win_rate",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "win_rate",
+          "unit": "ratio",
+          "value": 0.75
+        },
+        {
+          "availability": "available",
+          "key": "tracking_error",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "tracking_error",
+          "unit": "ratio_per_year",
+          "value": 1.3032171729991897
+        },
+        {
+          "availability": "available",
+          "key": "information_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "ir",
+          "unit": "ratio",
+          "value": 22.801264912443322
+        },
+        {
+          "availability": "available",
+          "key": "alpha",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "alpha",
+          "unit": "ratio_per_year",
+          "value": 123663320625.66454
+        },
+        {
+          "availability": "available",
+          "key": "beta",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "beta",
+          "unit": "ratio",
+          "value": 3.2500000000000013
+        },
+        {
+          "availability": "available",
+          "key": "trade_count",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "n_trades",
+          "unit": "count",
+          "value": 3
+        },
+        {
+          "availability": "available",
+          "key": "day_count",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "n_days",
+          "unit": "count",
+          "value": 4
+        }
+      ],
+      "performance_evidence_id": "rhperformanceevidencev1:sha256:f1f72c34d7427bd0d2e273a66b9677d112aa51c2417ea1cda8245e51e334e4f6",
+      "performance_row_digest": "sha256:bad2b506679a6ca13d80ab00859a442ba692f3872d42b0b45bbc3b1d6ffe72a4",
+      "performance_table_content_digest": "sha256:0856439ea7ec84e38887ccfa0067324f2e9cd293543e4ef683b9c2beb9eb34cf",
+      "performance_table_logical_name": "performance",
+      "performance_table_row_count": 1,
+      "performance_table_schema_digest": "sha256:16cef93a679761103ae405e622b7929abbfe07115bf276be4f164da5a16128d0",
+      "research_artifact_content_digest": "sha256:6ec566d73da3966d3d5f946313e2b0181993374c9f4ceb4b56e165cc6793d382",
+      "research_artifact_schema_version": "1.1.0",
+      "run_id": "rhbacktestrunv1:sha256:5036c771c44a2adade9ea590eee0d8cf824ff8f519fadd8ba424e5d914856386",
+      "schema_version": "researchhub.performance-evidence.v1",
+      "scope": "offline_research_only",
+      "start_date": "2026-01-05",
+      "strategy_digest": "sha256:6666666666666666666666666666666666666666666666666666666666666666",
+      "strategy_id": "alpha-top1",
+      "strategy_version": "1.0.0",
+      "timezone": "Asia/Shanghai"
+    },
+    "zero_active_variance": {
+      "artifact_available_at": "2026-01-08T02:05:00Z",
+      "authority": "quant_engine",
+      "backtest_evidence_manifest_document_sha256": "sha256:8a13212575154d97f85011ed54ed3bb588f73a691926b3ca03b360d563e9fc29",
+      "backtest_evidence_manifest_evidence_digest": "sha256:8c0c2052ab93aa920254bc7f4d89435d839b7150ecaaaa798703e89ddf876ec1",
+      "backtest_evidence_manifest_id": "rhbacktestevidencev1:sha256:699ad35ee30a294082c607c40617097aa90d5d58343e7d8c8185d8a53f52596c",
+      "backtest_evidence_qualification": "contract_qualified",
+      "backtest_run_ref_document_sha256": "sha256:6a798adb3e0568aca84ed3e3a285b92d181ec569462fc003952a8c0d8382ae3a",
+      "backtest_run_ref_id": "rhbacktestrunv1:sha256:5036c771c44a2adade9ea590eee0d8cf824ff8f519fadd8ba424e5d914856386",
+      "benchmark_alignment_policy": "exact_session_index",
+      "benchmark_id": "000300.SH",
+      "benchmark_series_digest": "sha256:7668b77ac689cd542005fae42cb2abd48fdf772d2f982086dca409606653e6ac",
+      "calendar": "CN-A",
+      "code_revision": "dddddddddddddddddddddddddddddddddddddddd",
+      "configuration_digest": "sha256:d28b49ea1bebc78d0023667f4fb9990b1fb45807176c2193b458525def056f4d",
+      "cost_model_digest": "sha256:8888888888888888888888888888888888888888888888888888888888888888",
+      "cost_model_version": "1.0.0",
+      "dataset_content_digest": "sha256:44ea11ba64dc2e6fd55c6d8e038c5edc84ee38d6fd46e38b15a1d5409662a020",
+      "dataset_manifest_digest": "sha256:d991bb2f8f6b80525f93c51e0b371213a3ed4649dffb073ed4605bfbd32349bd",
+      "dataset_snapshot_id": "rhdsv1:sha256:f63a29b4795c63fb7d6b2d3b5544cee9274b633db77c75c63d50a340c0827d57",
+      "document_sha256": "sha256:44a04be0105f29e1cbf8d2430b6b0eb084853d428bc2ccd5834dfdbcf9482bd7",
+      "end_date": "2026-01-08",
+      "environment_lock_digest": "sha256:9999999999999999999999999999999999999999999999999999999999999999",
+      "execution_model_digest": "sha256:7777777777777777777777777777777777777777777777777777777777777777",
+      "execution_model_version": "1.0.0",
+      "factor_output_content_digest": "sha256:d78751460dc27fc796163c462d946924ce2bcb45dcfceb5e4b77f76093befc9e",
+      "factor_set_digest": "sha256:e9339581cf569e92459f672e8081337712e7bf98e58ad42d60d7ed13f9b5a021",
+      "factor_set_id": "rhfactorsetv1:sha256:e9339581cf569e92459f672e8081337712e7bf98e58ad42d60d7ed13f9b5a021",
+      "foundation_digest": "sha256:d848237ab753ee9432ae78ec1f93b6ac45c8072d6694023b7f288203daf9d838",
+      "foundation_id": "rhdfv1:sha256:d848237ab753ee9432ae78ec1f93b6ac45c8072d6694023b7f288203daf9d838",
+      "frequency": "1d",
+      "methodology": {
+        "alpha": "daily_ols_intercept_geometric_annualization",
+        "annual_risk_free": 0.0,
+        "annualized_return": "geometric_compound",
+        "annualized_volatility": "sample_std_sqrt_periods",
+        "benchmark_alignment": "exact_session_index",
+        "benchmark_risk_free_daily": 0.0,
+        "beta": "sample_covariance_over_sample_variance",
+        "calmar_ratio": "unadjusted_annualized_return_over_absolute_maximum_drawdown",
+        "code_revision": "dddddddddddddddddddddddddddddddddddddddd",
+        "implementation_module": "quant_engine.metrics",
+        "implementation_version": "researchhub.quant-performance-methodology.v1",
+        "information_ratio": "mean_active_over_sample_std_active_sqrt_periods",
+        "maximum_drawdown": "non_positive_peak_to_trough_ratio_with_initial_nav_one",
+        "methodology_id": "researchhub.quant-performance-methodology.v1",
+        "periods_per_year": 252,
+        "return_type": "simple",
+        "sharpe_ratio": "annualized_return_minus_annual_risk_free_over_annualized_volatility",
+        "sortino_ratio": "annualized_return_minus_annual_risk_free_over_root_mean_square_negative_returns_sqrt_periods",
+        "source_frequency": "1d",
+        "total_return": "final_nav_minus_one",
+        "tracking_error": "sample_std_active_return_sqrt_periods",
+        "win_rate": "positive_daily_return_count_over_observation_count"
+      },
+      "methodology_id": "researchhub.quant-performance-methodology.v1",
+      "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+      "metrics": [
+        {
+          "availability": "available",
+          "key": "total_return",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "total_ret",
+          "unit": "ratio",
+          "value": 0.575
+        },
+        {
+          "availability": "available",
+          "key": "annualized_return",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "ann_ret",
+          "unit": "ratio_per_year",
+          "value": 2683336646708.1
+        },
+        {
+          "availability": "available",
+          "key": "annualized_volatility",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "ann_volatility",
+          "unit": "ratio_per_year",
+          "value": 1.38901943830891
+        },
+        {
+          "availability": "available",
+          "key": "sharpe_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "sharpe",
+          "unit": "ratio",
+          "value": 1931820803008.3313
+        },
+        {
+          "availability": "available",
+          "key": "sortino_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "sortino",
+          "unit": "ratio",
+          "value": 0.0
+        },
+        {
+          "availability": "available",
+          "key": "maximum_drawdown",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "max_dd",
+          "unit": "ratio",
+          "value": 0.0
+        },
+        {
+          "availability": "available",
+          "key": "calmar_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "calmar",
+          "unit": "ratio",
+          "value": 0.0
+        },
+        {
+          "availability": "available",
+          "key": "win_rate",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "win_rate",
+          "unit": "ratio",
+          "value": 0.75
+        },
+        {
+          "availability": "available",
+          "key": "tracking_error",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "tracking_error",
+          "unit": "ratio_per_year",
+          "value": 0.0
+        },
+        {
+          "availability": "not_estimable_active_variance",
+          "key": "information_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "ir",
+          "unit": "ratio",
+          "value": null
+        },
+        {
+          "availability": "available",
+          "key": "alpha",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "alpha",
+          "unit": "ratio_per_year",
+          "value": 0.0
+        },
+        {
+          "availability": "available",
+          "key": "beta",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "beta",
+          "unit": "ratio",
+          "value": 1.0000000000000002
+        },
+        {
+          "availability": "available",
+          "key": "trade_count",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "n_trades",
+          "unit": "count",
+          "value": 3
+        },
+        {
+          "availability": "available",
+          "key": "day_count",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "n_days",
+          "unit": "count",
+          "value": 4
+        }
+      ],
+      "performance_evidence_id": "rhperformanceevidencev1:sha256:e6730d25d90c85a6370adb72bc45c1bf505235f77b72521dd8197898ca369fe5",
+      "performance_row_digest": "sha256:6528c47fa9416e350aa5010477dbd8c83a35cecf87aa5bc413cec95a9765114d",
+      "performance_table_content_digest": "sha256:7d57a966431c68093a9f6193ae62f79a63d33832f11dd68524ed9bcd2cb6257c",
+      "performance_table_logical_name": "performance",
+      "performance_table_row_count": 1,
+      "performance_table_schema_digest": "sha256:16cef93a679761103ae405e622b7929abbfe07115bf276be4f164da5a16128d0",
+      "research_artifact_content_digest": "sha256:b17ab9e158a7ceeb5107f6fe8a61f332009c314186f3436d4750aa44d6ca3856",
+      "research_artifact_schema_version": "1.1.0",
+      "run_id": "rhbacktestrunv1:sha256:5036c771c44a2adade9ea590eee0d8cf824ff8f519fadd8ba424e5d914856386",
+      "schema_version": "researchhub.performance-evidence.v1",
+      "scope": "offline_research_only",
+      "start_date": "2026-01-05",
+      "strategy_digest": "sha256:6666666666666666666666666666666666666666666666666666666666666666",
+      "strategy_id": "alpha-top1",
+      "strategy_version": "1.0.0",
+      "timezone": "Asia/Shanghai"
+    },
+    "zero_benchmark_variance": {
+      "artifact_available_at": "2026-01-08T02:05:00Z",
+      "authority": "quant_engine",
+      "backtest_evidence_manifest_document_sha256": "sha256:2389b804f5d8d20a37b31f596b52892484777c5da799dd68865a082ee90e6e5b",
+      "backtest_evidence_manifest_evidence_digest": "sha256:54bb315bc1940ef6df79999cf03ebb8d10defaf526bf7802d63da33f612520ec",
+      "backtest_evidence_manifest_id": "rhbacktestevidencev1:sha256:ad9559e1f8feca28bb310765216a975935f05f51057139c2d02fd2fe7dfe501e",
+      "backtest_evidence_qualification": "contract_qualified",
+      "backtest_run_ref_document_sha256": "sha256:6a798adb3e0568aca84ed3e3a285b92d181ec569462fc003952a8c0d8382ae3a",
+      "backtest_run_ref_id": "rhbacktestrunv1:sha256:5036c771c44a2adade9ea590eee0d8cf824ff8f519fadd8ba424e5d914856386",
+      "benchmark_alignment_policy": "exact_session_index",
+      "benchmark_id": "000300.SH",
+      "benchmark_series_digest": "sha256:d0f1e954d796e95254bcccc893d6e3a8f9d541c3028b88b2c0b24c1658bb2408",
+      "calendar": "CN-A",
+      "code_revision": "dddddddddddddddddddddddddddddddddddddddd",
+      "configuration_digest": "sha256:d28b49ea1bebc78d0023667f4fb9990b1fb45807176c2193b458525def056f4d",
+      "cost_model_digest": "sha256:8888888888888888888888888888888888888888888888888888888888888888",
+      "cost_model_version": "1.0.0",
+      "dataset_content_digest": "sha256:44ea11ba64dc2e6fd55c6d8e038c5edc84ee38d6fd46e38b15a1d5409662a020",
+      "dataset_manifest_digest": "sha256:d991bb2f8f6b80525f93c51e0b371213a3ed4649dffb073ed4605bfbd32349bd",
+      "dataset_snapshot_id": "rhdsv1:sha256:f63a29b4795c63fb7d6b2d3b5544cee9274b633db77c75c63d50a340c0827d57",
+      "document_sha256": "sha256:077f3ec84802aff2f046365a3a5f61f475b1ad63ec05e96742844bc531bcb551",
+      "end_date": "2026-01-08",
+      "environment_lock_digest": "sha256:9999999999999999999999999999999999999999999999999999999999999999",
+      "execution_model_digest": "sha256:7777777777777777777777777777777777777777777777777777777777777777",
+      "execution_model_version": "1.0.0",
+      "factor_output_content_digest": "sha256:d78751460dc27fc796163c462d946924ce2bcb45dcfceb5e4b77f76093befc9e",
+      "factor_set_digest": "sha256:e9339581cf569e92459f672e8081337712e7bf98e58ad42d60d7ed13f9b5a021",
+      "factor_set_id": "rhfactorsetv1:sha256:e9339581cf569e92459f672e8081337712e7bf98e58ad42d60d7ed13f9b5a021",
+      "foundation_digest": "sha256:d848237ab753ee9432ae78ec1f93b6ac45c8072d6694023b7f288203daf9d838",
+      "foundation_id": "rhdfv1:sha256:d848237ab753ee9432ae78ec1f93b6ac45c8072d6694023b7f288203daf9d838",
+      "frequency": "1d",
+      "methodology": {
+        "alpha": "daily_ols_intercept_geometric_annualization",
+        "annual_risk_free": 0.0,
+        "annualized_return": "geometric_compound",
+        "annualized_volatility": "sample_std_sqrt_periods",
+        "benchmark_alignment": "exact_session_index",
+        "benchmark_risk_free_daily": 0.0,
+        "beta": "sample_covariance_over_sample_variance",
+        "calmar_ratio": "unadjusted_annualized_return_over_absolute_maximum_drawdown",
+        "code_revision": "dddddddddddddddddddddddddddddddddddddddd",
+        "implementation_module": "quant_engine.metrics",
+        "implementation_version": "researchhub.quant-performance-methodology.v1",
+        "information_ratio": "mean_active_over_sample_std_active_sqrt_periods",
+        "maximum_drawdown": "non_positive_peak_to_trough_ratio_with_initial_nav_one",
+        "methodology_id": "researchhub.quant-performance-methodology.v1",
+        "periods_per_year": 252,
+        "return_type": "simple",
+        "sharpe_ratio": "annualized_return_minus_annual_risk_free_over_annualized_volatility",
+        "sortino_ratio": "annualized_return_minus_annual_risk_free_over_root_mean_square_negative_returns_sqrt_periods",
+        "source_frequency": "1d",
+        "total_return": "final_nav_minus_one",
+        "tracking_error": "sample_std_active_return_sqrt_periods",
+        "win_rate": "positive_daily_return_count_over_observation_count"
+      },
+      "methodology_id": "researchhub.quant-performance-methodology.v1",
+      "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+      "metrics": [
+        {
+          "availability": "available",
+          "key": "total_return",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "total_ret",
+          "unit": "ratio",
+          "value": 0.575
+        },
+        {
+          "availability": "available",
+          "key": "annualized_return",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "ann_ret",
+          "unit": "ratio_per_year",
+          "value": 2683336646708.1
+        },
+        {
+          "availability": "available",
+          "key": "annualized_volatility",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "ann_volatility",
+          "unit": "ratio_per_year",
+          "value": 1.38901943830891
+        },
+        {
+          "availability": "available",
+          "key": "sharpe_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "sharpe",
+          "unit": "ratio",
+          "value": 1931820803008.3313
+        },
+        {
+          "availability": "available",
+          "key": "sortino_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "sortino",
+          "unit": "ratio",
+          "value": 0.0
+        },
+        {
+          "availability": "available",
+          "key": "maximum_drawdown",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "max_dd",
+          "unit": "ratio",
+          "value": 0.0
+        },
+        {
+          "availability": "available",
+          "key": "calmar_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "calmar",
+          "unit": "ratio",
+          "value": 0.0
+        },
+        {
+          "availability": "available",
+          "key": "win_rate",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "win_rate",
+          "unit": "ratio",
+          "value": 0.75
+        },
+        {
+          "availability": "available",
+          "key": "tracking_error",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "tracking_error",
+          "unit": "ratio_per_year",
+          "value": 1.38901943830891
+        },
+        {
+          "availability": "available",
+          "key": "information_ratio",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "ir",
+          "unit": "ratio",
+          "value": 22.299903907544408
+        },
+        {
+          "availability": "not_estimable_benchmark_variance",
+          "key": "alpha",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "alpha",
+          "unit": "ratio_per_year",
+          "value": null
+        },
+        {
+          "availability": "not_estimable_benchmark_variance",
+          "key": "beta",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": true,
+          "source_column": "beta",
+          "unit": "ratio",
+          "value": null
+        },
+        {
+          "availability": "available",
+          "key": "trade_count",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "n_trades",
+          "unit": "count",
+          "value": 3
+        },
+        {
+          "availability": "available",
+          "key": "day_count",
+          "methodology_id": "researchhub.quant-performance-methodology.v1",
+          "metric_schema_id": "researchhub.quant-performance-metrics.v1",
+          "nullable": false,
+          "source_column": "n_days",
+          "unit": "count",
+          "value": 4
+        }
+      ],
+      "performance_evidence_id": "rhperformanceevidencev1:sha256:86133a6b3c3091477cdc4618ebc294e671c9ba0a3ef5015a068f92bb575729b9",
+      "performance_row_digest": "sha256:90c9aa2f4168c15ff9ac5d4da2a35c3e915bb4e04736e5a440ea80c787a99553",
+      "performance_table_content_digest": "sha256:1570b64d83ad2a849607e6f5203f5ec2f0291a8d6cbb00b8edfe0ae030b7967d",
+      "performance_table_logical_name": "performance",
+      "performance_table_row_count": 1,
+      "performance_table_schema_digest": "sha256:16cef93a679761103ae405e622b7929abbfe07115bf276be4f164da5a16128d0",
+      "research_artifact_content_digest": "sha256:fed7a28888a6e98ccce78e7d84f22d4e3636e928485861af7a97f90eddc91f94",
+      "research_artifact_schema_version": "1.1.0",
+      "run_id": "rhbacktestrunv1:sha256:5036c771c44a2adade9ea590eee0d8cf824ff8f519fadd8ba424e5d914856386",
+      "schema_version": "researchhub.performance-evidence.v1",
+      "scope": "offline_research_only",
+      "start_date": "2026-01-05",
+      "strategy_digest": "sha256:6666666666666666666666666666666666666666666666666666666666666666",
+      "strategy_id": "alpha-top1",
+      "strategy_version": "1.0.0",
+      "timezone": "Asia/Shanghai"
+    }
+  },
+  "schema_version": 1,
+  "source_commit": "a724e1e57a99d1304a932d01ee836bac56c5c15c",
+  "source_tree": "4774e88442d25bf79a54eab3d7106ff4d0ba9603"
+}

tests/governance/test_module_spec.py

@@ -16,7 +16,7 @@ def test_module_spec_declares_pure_research_engine_boundary() -> None:
     prohibited = " ".join(spec["bounded_context"]["prohibited_responsibilities"]).lower()
     for term in ("investment advice", "live order", "credentials", "source facts"):
         assert term in prohibited
-    assert spec["authority"]["revision"] == 4
+    assert spec["authority"]["revision"] == 5
     assert {
         (item["contract_id"], item["version"])
         for item in spec["contracts"]["provides"]
@@ -25,6 +25,7 @@ def test_module_spec_declares_pure_research_engine_boundary() -> None:
         ("researchhub.factor-set-ref", "1.0.0"),
         ("researchhub.backtest-run-ref", "1.0.0"),
         ("researchhub.backtest-evidence-manifest", "1.0.0"),
+        ("researchhub.performance-evidence", "1.0.0"),
         ("researchhub.portfolio-decision", "1.0.0"),
         ("researchhub.risk-assessment", "1.0.0"),
     }
@@ -33,6 +34,7 @@ def test_module_spec_declares_pure_research_engine_boundary() -> None:
         "researchhub.factor-set-ref": "src/quant_engine/factor_contracts.py",
         "researchhub.backtest-run-ref": "src/quant_engine/governed_pipeline.py",
         "researchhub.backtest-evidence-manifest": "src/quant_engine/artifact.py",
+        "researchhub.performance-evidence": "src/quant_engine/artifact.py",
         "researchhub.portfolio-decision": "src/quant_engine/portfolio_risk_contracts.py",
         "researchhub.risk-assessment": "src/quant_engine/portfolio_risk_contracts.py",
     }
@@ -53,6 +55,11 @@ def test_module_spec_declares_pure_research_engine_boundary() -> None:
     )
     assert spec["dependencies"] == []
     capabilities = {item["id"]: item for item in spec["capabilities"]}
+    evidence_contract = capabilities["backtest-evidence-contracts"]
+    assert evidence_contract["status"] == "operational"
+    evidence_summary = evidence_contract["summary"].lower()
+    for term in ("performance-methodology", "without recomputation", "decision authority"):
+        assert term in evidence_summary
     portfolio_contract = capabilities["portfolio-risk-computation-contracts"]
     assert portfolio_contract["status"] == "operational"
     summary = portfolio_contract["summary"].lower()

tests/test_performance_evidence_contract.py

@@ -0,0 +1,727 @@
+"""Closed performance-evidence contract conformance tests."""
+
+from __future__ import annotations
+
+import copy
+import hashlib
+import json
+from dataclasses import replace
+from pathlib import Path
+from typing import Any
+
+import numpy as np
+import pandas as pd
+import pytest
+
+from quant_engine.artifact import (
+    PERFORMANCE_EVIDENCE_SCHEMA_VERSION,
+    PERFORMANCE_METRIC_SCHEMA_ID,
+    PERFORMANCE_METHODOLOGY_ID,
+    BacktestEvidenceManifest,
+    EvidenceQualification,
+    PerformanceEvidenceError,
+    PerformanceEvidenceErrorCode,
+    PerformanceEvidenceV1,
+    PerformanceMetricAvailability,
+    ResearchRunArtifact,
+    build_backtest_evidence_manifest,
+    build_performance_evidence,
+    build_research_run_artifact,
+)
+from quant_engine.execution import ExecutionConfig
+from quant_engine.factor_contracts import (
+    ActorIdentity,
+    AvailabilityMode,
+    Causation,
+    DataFoundationEnvelope,
+    DatasetSnapshotEnvelope,
+    FactorInput,
+    FactorSetRef,
+    InputBinding,
+    OutputArtifactRef,
+    OutputCoverage,
+    OutputQuality,
+    OutputQualityCheck,
+    ProducerIdentity,
+    ViewAvailability,
+    canonical_json_bytes,
+    factor_definition_from_alpha158,
+    factor_input_schema_digest,
+)
+from quant_engine.governed_pipeline import BacktestRunRef
+from quant_engine.metrics import TRADING_DAYS_PER_YEAR, benchmark_summary, summary
+from quant_engine.research_pipeline import FactorBacktestResult, run_factor_backtest_research
+
+
+ROOT = Path(__file__).resolve().parents[1]
+FACTOR_FIXTURE = ROOT / "tests" / "fixtures" / "factor-contracts-v1.golden.json"
+PERFORMANCE_FIXTURE = (
+    ROOT / "tests" / "fixtures" / "performance-evidence-v1.golden.json"
+)
+VIEW_REF_ID = "rhviewrefv1:sha256:bf776bcd26d940fafde1d650776a5505fb3fe8b5b068c351622bf2c42385629c"
+VIEW_SCHEMA_DIGEST = "sha256:0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef"
+CALENDAR_REVISION_ID = "rhcalv1:sha256:1f4ca22557063389badd669cf774bb35234066e847646682dccfe411e252a078"
+ACTION_REVISION_ID = "rhcav1:sha256:0f947df29f152bfa2c6ab0da7a0d464670c7ad2526cd10f2a7939b42fee5c275"
+PARAMETERS = {"lag_sessions": 1, "top_k": 1}
+
+
+def _sha256(value: bytes) -> str:
+    return f"sha256:{hashlib.sha256(value).hexdigest()}"
+
+
+def _accepted_authorities() -> tuple[
+    DatasetSnapshotEnvelope,
+    DataFoundationEnvelope,
+    FactorSetRef,
+]:
+    fixture = json.loads(FACTOR_FIXTURE.read_text(encoding="utf-8"))
+    snapshot = DatasetSnapshotEnvelope.from_dict(fixture["dataset_snapshot"])
+    foundation = DataFoundationEnvelope.from_dict(fixture["data_foundation"])
+    factor_input = FactorInput("market", VIEW_SCHEMA_DIGEST, ("close", "volume"))
+    definition = factor_definition_from_alpha158(
+        "alpha_005",
+        version="1.0.0",
+        parameters={},
+        inputs=(factor_input,),
+        implementation_digest="sha256:" + "1" * 64,
+        input_schema_digest=factor_input_schema_digest((factor_input,)),
+        valid_from="2026-01-01T00:00:00.000000Z",
+        valid_until="2027-01-01T00:00:00Z",
+        warmup_sessions=10,
+        lag_sessions=1,
+        producer=ProducerIdentity("quant_engine", "1.0.0"),
+        code_revision="c" * 40,
+    )
+    output_schema_bytes = canonical_json_bytes(fixture["output_schema"])
+    output_content_bytes = canonical_json_bytes(fixture["output_content"])
+    artifact_ref = OutputArtifactRef.create(
+        schema_digest=_sha256(output_schema_bytes),
+        content_digest=_sha256(output_content_bytes),
+    )
+    factor_set = FactorSetRef.create(
+        definitions=(definition,),
+        dataset_snapshot=snapshot,
+        foundation=foundation,
+        selected_view_ref_ids=(VIEW_REF_ID,),
+        input_bindings=(
+            InputBinding(
+                definition.definition_id,
+                "market",
+                VIEW_REF_ID,
+                VIEW_SCHEMA_DIGEST,
+            ),
+        ),
+        view_availability=(
+            ViewAvailability(VIEW_REF_ID, "2026-01-02T23:50:00Z", "sha256:" + "2" * 64),
+        ),
+        output_quality=OutputQuality(
+            "passed",
+            (OutputQualityCheck("finite_values", "passed", "sha256:" + "3" * 64),),
+        ),
+        output_coverage=OutputCoverage(
+            "complete",
+            1,
+            1,
+            "row",
+            "alpha_005.cn_a",
+            "sha256:" + "4" * 64,
+        ),
+        output_schema_bytes=output_schema_bytes,
+        output_content_bytes=output_content_bytes,
+        output_artifact_ref=artifact_ref,
+        availability_mode=AvailabilityMode.AS_AVAILABLE,
+        evaluation_at="2026-01-03T11:00:00Z",
+        computed_at="2026-01-03T10:15:00Z",
+        artifact_available_at="2026-01-03T10:20:00Z",
+        producer=ProducerIdentity("quant_engine", "1.0.0"),
+        code_revision="c" * 40,
+        actor=ActorIdentity("service", "factor_worker_v1"),
+        correlation_id="research_run_001",
+        causation=Causation("foundation", foundation.foundation_id),
+        evidence_scope="synthetic_fixture",
+        decision_eligible=False,
+    )
+    return snapshot, foundation, factor_set
+
+
+def _configuration_digest() -> str:
+    return _sha256(
+        json.dumps(
+            PARAMETERS,
+            ensure_ascii=False,
+            sort_keys=True,
+            separators=(",", ":"),
+            allow_nan=False,
+        ).encode("utf-8")
+    )
+
+
+def _run_ref(**overrides: Any) -> BacktestRunRef:
+    snapshot, foundation, factor_set = _accepted_authorities()
+    arguments: dict[str, Any] = {
+        "dataset_snapshot": snapshot,
+        "foundation": foundation,
+        "factor_set": factor_set,
+        "universe_digest": "sha256:" + "5" * 64,
+        "trading_calendar_revision_ids": (CALENDAR_REVISION_ID,),
+        "corporate_action_revision_ids": (ACTION_REVISION_ID,),
+        "strategy_id": "alpha-top1",
+        "strategy_version": "1.0.0",
+        "strategy_digest": "sha256:" + "6" * 64,
+        "execution_model_version": "1.0.0",
+        "execution_model_digest": "sha256:" + "7" * 64,
+        "cost_model_version": "1.0.0",
+        "cost_model_digest": "sha256:" + "8" * 64,
+        "random_seed": 7,
+        "code_revision": "d" * 40,
+        "environment_lock_digest": "sha256:" + "9" * 64,
+        "configuration_digest": _configuration_digest(),
+        "evaluation_at": "2026-01-08T01:00:00Z",
+        "computed_at": "2026-01-08T02:00:00Z",
+    }
+    arguments.update(overrides)
+    return BacktestRunRef.create(**arguments)
+
+
+def _backtest_result() -> FactorBacktestResult:
+    dates = pd.date_range("2026-01-05", periods=4, freq="B")
+    scores = pd.DataFrame({"A": [2.0, 0.0], "B": [1.0, 3.0]}, index=dates[:2])
+    opens = pd.DataFrame(
+        {"A": [10.0, 10.0, 15.0, 15.0], "B": [20.0, 20.0, 20.0, 21.0]},
+        index=dates,
+    )
+    closes = pd.DataFrame(
+        {"A": [10.0, 12.0, 15.0, 15.0], "B": [20.0, 20.0, 18.0, 21.0]},
+        index=dates,
+    )
+    return run_factor_backtest_research(
+        scores,
+        opens,
+        closes,
+        top_k=1,
+        execution_price_field="open",
+        valuation_price_field="close",
+        initial_cash=1_000.0,
+        config=ExecutionConfig(
+            commission_bps=0,
+            stamp_tax_bps=0,
+            slippage_bps=0,
+            min_trade_amount=0,
+        ),
+    )
+
+
+def _artifact(
+    run_ref: BacktestRunRef,
+    benchmark_kind: str,
+) -> tuple[ResearchRunArtifact, FactorBacktestResult]:
+    result = _backtest_result()
+    benchmark_id: str | None
+    benchmark_returns: pd.Series | None
+    if benchmark_kind == "absent":
+        benchmark_id = None
+        benchmark_returns = None
+    elif benchmark_kind == "estimable":
+        benchmark_id = "000300.SH"
+        benchmark_returns = pd.Series(
+            [0.0, 0.01, -0.01, 0.02],
+            index=result.returns.index,
+            name="benchmark_return",
+        )
+    elif benchmark_kind == "zero_active_variance":
+        benchmark_id = "000300.SH"
+        benchmark_returns = result.returns.rename("benchmark_return")
+    elif benchmark_kind == "zero_benchmark_variance":
+        benchmark_id = "000300.SH"
+        benchmark_returns = pd.Series(
+            np.zeros(len(result.returns)),
+            index=result.returns.index,
+            name="benchmark_return",
+        )
+    else:
+        raise AssertionError(f"unknown benchmark_kind: {benchmark_kind}")
+    artifact = build_research_run_artifact(
+        result,
+        run_id=run_ref.run_id,
+        strategy_id=run_ref.strategy_id,
+        strategy_name="Alpha Top 1",
+        strategy_version=run_ref.strategy_version,
+        engine_version="1.2.0",
+        code_revision=run_ref.code_revision,
+        data_snapshot_id=run_ref.dataset_snapshot_id,
+        calendar="CN-A",
+        timezone="Asia/Shanghai",
+        started_at="2026-01-08T10:00:00+08:00",
+        finished_at="2026-01-08T10:01:00+08:00",
+        parameters=PARAMETERS,
+        benchmark_id=benchmark_id,
+        benchmark_returns=benchmark_returns,
+    )
+    return artifact, result
+
+
+def _case(
+    benchmark_kind: str,
+) -> tuple[PerformanceEvidenceV1, ResearchRunArtifact, BacktestRunRef, BacktestEvidenceManifest]:
+    run_ref = _run_ref()
+    artifact, _ = _artifact(run_ref, benchmark_kind)
+    manifest = build_backtest_evidence_manifest(
+        run_ref,
+        artifact,
+        artifact_available_at="2026-01-08T02:05:00Z",
+        qualification=EvidenceQualification.CONTRACT_QUALIFIED,
+    )
+    return (
+        build_performance_evidence(artifact, run_ref, manifest),
+        artifact,
+        run_ref,
+        manifest,
+    )
+
+
+def _metric_map(evidence: PerformanceEvidenceV1) -> dict[str, Any]:
+    return {metric.key: metric for metric in evidence.metrics}
+
+
+def _mutate_frozen(value: Any, field: str, replacement: object) -> Any:
+    changed = copy.copy(value)
+    object.__setattr__(changed, field, replacement)
+    return changed
+
+
+def _assert_error(
+    error: pytest.ExceptionInfo[PerformanceEvidenceError],
+    code: PerformanceEvidenceErrorCode,
+    path: str,
+) -> None:
+    assert error.value.code is code
+    assert error.value.path == path
+
+
+def test_present_evidence_is_deterministic_content_addressed_and_three_party_closed() -> None:
+    first, artifact, run_ref, manifest = _case("estimable")
+    second = build_performance_evidence(artifact, run_ref, manifest)
+
+    assert first == second
+    assert first.schema_version == PERFORMANCE_EVIDENCE_SCHEMA_VERSION
+    assert first.performance_evidence_id.startswith("rhperformanceevidencev1:sha256:")
+    assert first.document_sha256.startswith("sha256:")
+    assert first.authority == "quant_engine"
+    assert first.scope == "offline_research_only"
+    assert first.run_id == first.backtest_run_ref_id == run_ref.run_id == manifest.run_id
+    assert first.backtest_evidence_manifest_id == manifest.manifest_id
+    assert first.backtest_evidence_manifest_evidence_digest == manifest.evidence_digest
+    assert first.backtest_evidence_qualification == "contract_qualified"
+    assert first.research_artifact_content_digest == f"sha256:{artifact.content_sha256}"
+    assert first.performance_table_logical_name == "performance"
+    assert first.performance_table_row_count == 1
+    assert first.performance_row_digest.startswith("sha256:")
+    assert first.benchmark_series_digest is not None
+    assert first.canonical_bytes() == first.to_json().encode("utf-8")
+    assert not first.canonical_bytes().endswith(b"\n")
+    document_payload = first.to_dict()
+    document_payload.pop("document_sha256")
+    expected_document = json.dumps(
+        document_payload,
+        ensure_ascii=False,
+        sort_keys=True,
+        separators=(",", ":"),
+        allow_nan=False,
+    ).encode("utf-8")
+    assert _sha256(expected_document) == first.document_sha256
+    assert PerformanceEvidenceV1.from_dict(
+        first.to_dict(),
+        artifact=artifact,
+        run_ref=run_ref,
+        evidence_manifest=manifest,
+    ) == first
+
+
+def test_methodology_and_metrics_bind_the_actual_artifact_builder_path() -> None:
+    evidence, artifact, _, _ = _case("estimable")
+    result = _backtest_result()
+    expected_absolute = summary(result.returns, rf=0.0)
+    benchmark = artifact.nav.set_index("trade_date")["benchmark_return"]
+    benchmark.index = result.returns.index
+    expected_relative = benchmark_summary(
+        result.returns,
+        benchmark,
+        risk_free_daily=0.0,
+        annualization=TRADING_DAYS_PER_YEAR,
+    )
+    metrics = _metric_map(evidence)
+
+    assert evidence.methodology.methodology_id == PERFORMANCE_METHODOLOGY_ID
+    assert evidence.metric_schema_id == PERFORMANCE_METRIC_SCHEMA_ID
+    assert evidence.methodology.return_type == "simple"
+    assert evidence.methodology.source_frequency == "1d"
+    assert evidence.methodology.periods_per_year == TRADING_DAYS_PER_YEAR == 252
+    assert evidence.methodology.annual_risk_free == 0.0
+    assert evidence.methodology.benchmark_risk_free_daily == 0.0
+    assert evidence.methodology.benchmark_alignment == "exact_session_index"
+    assert metrics["annualized_return"].value == pytest.approx(
+        expected_absolute["ann_return"]
+    )
+    assert metrics["sharpe_ratio"].value == pytest.approx(expected_absolute["sharpe"])
+    assert metrics["tracking_error"].value == pytest.approx(
+        expected_relative["tracking_error"]
+    )
+    assert metrics["alpha"].value == pytest.approx(expected_relative["alpha"])
+    assert all(metric.methodology_id == PERFORMANCE_METHODOLOGY_ID for metric in metrics.values())
+    assert all(metric.metric_schema_id == PERFORMANCE_METRIC_SCHEMA_ID for metric in metrics.values())
+
+
+def test_relative_metric_availability_is_closed_for_present_absent_and_unestimable() -> None:
+    present, *_ = _case("estimable")
+    absent, *_ = _case("absent")
+    zero_active, *_ = _case("zero_active_variance")
+    zero_benchmark, *_ = _case("zero_benchmark_variance")
+
+    present_metrics = _metric_map(present)
+    assert all(
+        present_metrics[key].availability is PerformanceMetricAvailability.AVAILABLE
+        for key in ("tracking_error", "information_ratio", "alpha", "beta")
+    )
+    absent_metrics = _metric_map(absent)
+    assert absent.benchmark_series_digest is None
+    assert absent.benchmark_id == ""
+    assert absent.benchmark_alignment_policy == "none"
+    assert all(
+        absent_metrics[key].value is None
+        and absent_metrics[key].availability
+        is PerformanceMetricAvailability.BENCHMARK_ABSENT
+        for key in ("tracking_error", "information_ratio", "alpha", "beta")
+    )
+    zero_active_metrics = _metric_map(zero_active)
+    assert zero_active_metrics["tracking_error"].value == pytest.approx(0.0)
+    assert (
+        zero_active_metrics["information_ratio"].availability
+        is PerformanceMetricAvailability.NOT_ESTIMABLE_ACTIVE_VARIANCE
+    )
+    assert zero_active_metrics["information_ratio"].value is None
+    zero_benchmark_metrics = _metric_map(zero_benchmark)
+    assert np.isfinite(zero_benchmark_metrics["tracking_error"].value)
+    for key in ("alpha", "beta"):
+        assert zero_benchmark_metrics[key].value is None
+        assert (
+            zero_benchmark_metrics[key].availability
+            is PerformanceMetricAvailability.NOT_ESTIMABLE_BENCHMARK_VARIANCE
+        )
+
+
+def test_golden_covers_present_absent_and_both_unestimable_states() -> None:
+    expected = {
+        "schema_version": 1,
+        "source_commit": "a724e1e57a99d1304a932d01ee836bac56c5c15c",
+        "source_tree": "4774e88442d25bf79a54eab3d7106ff4d0ba9603",
+        "cases": {
+            name: _case(name)[0].to_dict()
+            for name in (
+                "estimable",
+                "zero_active_variance",
+                "zero_benchmark_variance",
+                "absent",
+            )
+        },
+    }
+    assert json.loads(PERFORMANCE_FIXTURE.read_text(encoding="utf-8")) == expected
+
+
+@pytest.mark.parametrize(
+    ("owner", "field", "replacement", "code", "path"),
+    [
+        (
+            "run_ref",
+            "run_id",
+            "rhbacktestrunv1:sha256:" + "0" * 64,
+            PerformanceEvidenceErrorCode.IDENTITY_MISMATCH,
+            "$.backtest_run_ref.run_id",
+        ),
+        (
+            "manifest",
+            "manifest_id",
+            "rhbacktestevidencev1:sha256:" + "0" * 64,
+            PerformanceEvidenceErrorCode.EVIDENCE_MISMATCH,
+            "$.backtest_evidence_manifest.manifest_id",
+        ),
+        (
+            "manifest",
+            "qualification",
+            EvidenceQualification.EXPLORATORY,
+            PerformanceEvidenceErrorCode.AUTHORITY_REJECTED,
+            "$.backtest_evidence_manifest.qualification",
+        ),
+    ],
+)
+def test_owner_identity_and_authority_mismatches_fail_closed(
+    owner: str,
+    field: str,
+    replacement: object,
+    code: PerformanceEvidenceErrorCode,
+    path: str,
+) -> None:
+    _, artifact, run_ref, manifest = _case("estimable")
+    changed_run_ref = _mutate_frozen(run_ref, field, replacement) if owner == "run_ref" else run_ref
+    changed_manifest = (
+        _mutate_frozen(manifest, field, replacement) if owner == "manifest" else manifest
+    )
+    with pytest.raises(PerformanceEvidenceError) as rejected:
+        build_performance_evidence(artifact, changed_run_ref, changed_manifest)
+    _assert_error(rejected, code, path)
+
+
+def test_performance_table_row_and_benchmark_digest_mismatches_fail_closed() -> None:
+    evidence, artifact, run_ref, manifest = _case("estimable")
+    performance = artifact.performance
+    performance.loc[0, "n_days"] += 1
+    changed_artifact = replace(artifact, _performance=performance)
+    with pytest.raises(PerformanceEvidenceError) as table_mismatch:
+        build_performance_evidence(changed_artifact, run_ref, manifest)
+    _assert_error(
+        table_mismatch,
+        PerformanceEvidenceErrorCode.EVIDENCE_MISMATCH,
+        "$.artifact.tables.performance.content_digest",
+    )
+
+    payload = evidence.to_dict()
+    payload["benchmark_series_digest"] = "sha256:" + "0" * 64
+    with pytest.raises(PerformanceEvidenceError) as benchmark_mismatch:
+        PerformanceEvidenceV1.from_dict(
+            payload,
+            artifact=artifact,
+            run_ref=run_ref,
+            evidence_manifest=manifest,
+        )
+    _assert_error(
+        benchmark_mismatch,
+        PerformanceEvidenceErrorCode.BENCHMARK_INVALID,
+        "$.benchmark_series_digest",
+    )
+
+
+def test_manifest_closure_covers_non_performance_artifact_tables() -> None:
+    _, artifact, run_ref, manifest = _case("estimable")
+    nav = artifact.nav
+    nav.loc[0, "nav"] += 0.01
+    changed_artifact = replace(artifact, _nav=nav)
+
+    with pytest.raises(PerformanceEvidenceError) as rejected:
+        build_performance_evidence(changed_artifact, run_ref, manifest)
+
+    _assert_error(
+        rejected,
+        PerformanceEvidenceErrorCode.EVIDENCE_MISMATCH,
+        "$.artifact.tables.nav.content_digest",
+    )
+
+
+def test_row_and_benchmark_mutations_change_their_digests_and_document_identity() -> None:
+    original, artifact, run_ref, _ = _case("estimable")
+    performance = artifact.performance
+    performance.loc[0, "sharpe"] += 0.01
+    changed_performance_artifact = replace(artifact, _performance=performance)
+    changed_performance_manifest = build_backtest_evidence_manifest(
+        run_ref,
+        changed_performance_artifact,
+        artifact_available_at="2026-01-08T02:05:00Z",
+    )
+    changed_performance = build_performance_evidence(
+        changed_performance_artifact,
+        run_ref,
+        changed_performance_manifest,
+    )
+    assert changed_performance.performance_row_digest != original.performance_row_digest
+    assert changed_performance.performance_evidence_id != original.performance_evidence_id
+
+    nav = artifact.nav
+    nav.loc[0, "benchmark_nav"] += 0.01
+    changed_benchmark_artifact = replace(artifact, _nav=nav)
+    changed_benchmark_manifest = build_backtest_evidence_manifest(
+        run_ref,
+        changed_benchmark_artifact,
+        artifact_available_at="2026-01-08T02:05:00Z",
+    )
+    changed_benchmark = build_performance_evidence(
+        changed_benchmark_artifact,
+        run_ref,
+        changed_benchmark_manifest,
+    )
+    assert changed_benchmark.benchmark_series_digest != original.benchmark_series_digest
+    assert changed_benchmark.performance_row_digest == original.performance_row_digest
+    assert changed_benchmark.performance_evidence_id != original.performance_evidence_id
+
+
+def test_relative_metric_null_reasons_cannot_be_invented() -> None:
+    _, artifact, run_ref, _ = _case("estimable")
+    performance = artifact.performance
+    performance.loc[0, "alpha"] = float("nan")
+    changed_artifact = replace(artifact, _performance=performance)
+    changed_manifest = build_backtest_evidence_manifest(
+        run_ref,
+        changed_artifact,
+        artifact_available_at="2026-01-08T02:05:00Z",
+    )
+    with pytest.raises(PerformanceEvidenceError) as false_alpha_domain:
+        build_performance_evidence(changed_artifact, run_ref, changed_manifest)
+    _assert_error(
+        false_alpha_domain,
+        PerformanceEvidenceErrorCode.METRIC_INVALID,
+        "$.metrics.alpha.value",
+    )
+
+    _, absent_artifact, absent_run_ref, _ = _case("absent")
+    absent_performance = absent_artifact.performance
+    absent_performance.loc[0, "tracking_error"] = 0.0
+    changed_absent = replace(absent_artifact, _performance=absent_performance)
+    changed_absent_manifest = build_backtest_evidence_manifest(
+        absent_run_ref,
+        changed_absent,
+        artifact_available_at="2026-01-08T02:05:00Z",
+    )
+    with pytest.raises(PerformanceEvidenceError) as false_absence:
+        build_performance_evidence(
+            changed_absent,
+            absent_run_ref,
+            changed_absent_manifest,
+        )
+    _assert_error(
+        false_absence,
+        PerformanceEvidenceErrorCode.BENCHMARK_INVALID,
+        "$.metrics.tracking_error.availability",
+    )
+
+
+def test_artifact_builder_enforces_strict_benchmark_session_alignment() -> None:
+    run_ref = _run_ref()
+    result = _backtest_result()
+    misaligned = pd.Series(
+        [0.0, 0.01, -0.01, 0.02],
+        index=result.returns.index.shift(1, freq="B"),
+    )
+    with pytest.raises(ValueError, match="matching indexes"):
+        build_research_run_artifact(
+            result,
+            run_id=run_ref.run_id,
+            strategy_id=run_ref.strategy_id,
+            strategy_name="Alpha Top 1",
+            strategy_version=run_ref.strategy_version,
+            engine_version="1.2.0",
+            code_revision=run_ref.code_revision,
+            data_snapshot_id=run_ref.dataset_snapshot_id,
+            calendar="CN-A",
+            timezone="Asia/Shanghai",
+            started_at="2026-01-08T10:00:00+08:00",
+            finished_at="2026-01-08T10:01:00+08:00",
+            parameters=PARAMETERS,
+            benchmark_id="000300.SH",
+            benchmark_returns=misaligned,
+        )
+
+
+@pytest.mark.parametrize(
+    ("column", "value", "path"),
+    [
+        ("total_ret", -1.01, "$.metrics.total_return.value"),
+        ("ann_ret", -1.01, "$.metrics.annualized_return.value"),
+        ("ann_volatility", -0.01, "$.metrics.annualized_volatility.value"),
+        ("max_dd", 0.01, "$.metrics.maximum_drawdown.value"),
+        ("win_rate", 1.01, "$.metrics.win_rate.value"),
+        ("tracking_error", -0.01, "$.metrics.tracking_error.value"),
+        ("n_trades", True, "$.metrics.trade_count.value"),
+    ],
+)
+def test_metric_domains_reject_invalid_source_values(
+    column: str,
+    value: object,
+    path: str,
+) -> None:
+    _, artifact, run_ref, _ = _case("estimable")
+    performance = artifact.performance.astype(object)
+    performance.at[0, column] = value
+    changed_artifact = replace(artifact, _performance=performance)
+    changed_manifest = build_backtest_evidence_manifest(
+        run_ref,
+        changed_artifact,
+        artifact_available_at="2026-01-08T02:05:00Z",
+    )
+    with pytest.raises(PerformanceEvidenceError) as rejected:
+        build_performance_evidence(changed_artifact, run_ref, changed_manifest)
+    _assert_error(rejected, PerformanceEvidenceErrorCode.METRIC_INVALID, path)
+
+
+def test_closed_parser_rejects_unknown_non_ascii_non_finite_bool_and_unsafe_integer() -> None:
+    evidence, artifact, run_ref, manifest = _case("estimable")
+
+    mutations: list[tuple[dict[str, Any], PerformanceEvidenceErrorCode, str]] = []
+    unknown = evidence.to_dict()
+    unknown["unexpected"] = "value"
+    mutations.append((unknown, PerformanceEvidenceErrorCode.TYPE_ERROR, "$.unexpected"))
+    non_ascii = evidence.to_dict()
+    non_ascii["métric"] = "value"
+    mutations.append((non_ascii, PerformanceEvidenceErrorCode.TYPE_ERROR, "$.métric"))
+    non_finite = evidence.to_dict()
+    non_finite["metrics"][0]["value"] = float("inf")
+    mutations.append(
+        (non_finite, PerformanceEvidenceErrorCode.METRIC_INVALID, "$.metrics[0].value")
+    )
+    bool_number = evidence.to_dict()
+    bool_number["methodology"]["periods_per_year"] = True
+    mutations.append(
+        (
+            bool_number,
+            PerformanceEvidenceErrorCode.METHODOLOGY_MISMATCH,
+            "$.methodology.periods_per_year",
+        )
+    )
+    unsafe = evidence.to_dict()
+    unsafe["performance_table_row_count"] = 2**53
+    mutations.append(
+        (
+            unsafe,
+            PerformanceEvidenceErrorCode.EVIDENCE_MISMATCH,
+            "$.performance_table_row_count",
+        )
+    )
+
+    for payload, code, path in mutations:
+        with pytest.raises(PerformanceEvidenceError) as rejected:
+            PerformanceEvidenceV1.from_dict(
+                payload,
+                artifact=artifact,
+                run_ref=run_ref,
+                evidence_manifest=manifest,
+            )
+        _assert_error(rejected, code, path)
+
+
+def test_public_mapping_has_no_raw_inputs_storage_or_runtime_authority() -> None:
+    evidence, *_ = _case("estimable")
+    payload = evidence.to_dict()
+    serialized = evidence.to_json().lower()
+    forbidden_keys = {
+        "parameters",
+        "params_json",
+        "returns",
+        "nav",
+        "benchmark_series",
+        "table_bytes",
+        "locator",
+        "uri",
+        "credential",
+        "decision_eligible",
+        "publication_eligible",
+        "paper_trading",
+        "live_trading",
+        "investment_advice",
+    }
+
+    def keys(value: object) -> set[str]:
+        if isinstance(value, dict):
+            return set(value) | {key for item in value.values() for key in keys(item)}
+        if isinstance(value, list):
+            return {key for item in value for key in keys(item)}
+        return set()
+
+    assert not (keys(payload) & forbidden_keys)
+    for token in ("postgres://", "mysql://", "s3://", "credential", "broker"):
+        assert token not in serialized

tests/test_portfolio_risk_contracts.py

@@ -952,6 +952,37 @@ def test_architecture_dependency_no_copy_and_authority_boundaries() -> None:
     assert not any(name.startswith(("research_results", "research_platform")) for name in imports)
     for candidate in ("riskfolio", "pyp", "skfolio", "cvxportfolio"):
         assert candidate not in source.lower()
+    artifact_source = (ROOT / "src" / "quant_engine" / "artifact.py").read_text(
+        encoding="utf-8"
+    )
+    artifact_tree = ast.parse(artifact_source)
+    forbidden_artifact_authority_symbols = {
+        "PortfolioDecision",
+        "RiskAssessment",
+        "build_portfolio_decision",
+        "assess_portfolio_risk",
+    }
+    artifact_imports = {
+        alias.name
+        for node in ast.walk(artifact_tree)
+        if isinstance(node, ast.Import)
+        for alias in node.names
+    } | {
+        node.module or ""
+        for node in ast.walk(artifact_tree)
+        if isinstance(node, ast.ImportFrom)
+    }
+    artifact_names = {
+        node.id for node in ast.walk(artifact_tree) if isinstance(node, ast.Name)
+    } | {
+        node.attr for node in ast.walk(artifact_tree) if isinstance(node, ast.Attribute)
+    }
+    assert "quant_engine.portfolio_risk_contracts" not in artifact_imports
+    assert not forbidden_artifact_authority_symbols & artifact_names
+    assert all(
+        token not in artifact_source
+        for token in {"portfolio_risk_contracts", *forbidden_artifact_authority_symbols}
+    )
     for owner_path in (
         ROOT / "src" / "quant_engine" / "governed_pipeline.py",
         ROOT / "src" / "quant_engine" / "artifact.py",
@@ -989,7 +1020,6 @@ def test_architecture_dependency_no_copy_and_authority_boundaries() -> None:
 def test_read_only_owner_dependency_lock_and_ci_hashes_match_baseline() -> None:
     expected = {
         "src/quant_engine/governed_pipeline.py": "3b334f340898db78ed869375ab532f156e8c1fee595a8318f544bebdd391049d",
-        "src/quant_engine/artifact.py": "e15feec412d3bfff10d8f21ca20813fc65cabc0703940371ed661896147bc379",
         "src/quant_engine/portfolio_construction.py": "e93d71da8d61b2047c19d4b99dace934a8cbc96d8d2b150ad62a9ceebd9163d4",
         "src/quant_engine/portfolio_decomp.py": "1a4f9f9aac2c46bf6ed2826d1b0d1723f06e3ce7c4b3f6479e098cbbd135bea6",
         "src/quant_engine/risk.py": "4a66c312d517d40f6f67bb71f438523e135645624ae78d49fa9c0fda2c02074e",