cschanhniem · Jul 1, 2026 · Jul 1, 2026
diff --git a/Makefile b/Makefile
@@ -260,6 +260,10 @@ wave0-5-panel: wave0-5-fill-external
 wave0-5-evidence:
 	PYTHONPATH=src $(PYTHON) scripts/generate_wave0_5_evidence_certs.py
 
+metrics-snapshot:
+	PYTHONPATH=src $(PYTHON) -m openamp_foundry.cli bench metrics-snapshot \
+		--out outputs/metrics_snapshot.json
+
 wave0-5b-generate:
 	PYTHONPATH=src $(PYTHON) scripts/generate_wave0_5b_candidates.py
 

diff --git a/docs/BENCHMARK_CARD.md b/docs/BENCHMARK_CARD.md
@@ -109,9 +109,9 @@ against the simple ensemble to test whether its additional complexity earns its
 | Scorer | AUROC | CI₉₅ | Delta vs ensemble |
 |--------|:-----:|:----:|:-----------------:|
 | **Ensemble** (pipeline.yaml) | **0.7832** | 0.717–0.8423 | — |
-| Expert composite | 0.7360 | 0.6604–0.8037 | **−0.0472** |
+| Expert composite | 0.7119 | 0.6295–0.7865 | **−0.0713** |
 | **Ensemble** (phase3.yaml) | **0.7448** | 0.6741–0.8118 | — |
-| Expert composite | 0.7360 | 0.6604–0.8037 | −0.0088 |
+| Expert composite | 0.7119 | 0.6295–0.7865 | −0.0329 |
 
 **Per-component AUROC:**
 

diff --git a/docs/METRICS_CURRENT.md b/docs/METRICS_CURRENT.md
@@ -1,5 +1,7 @@
 # Current Pipeline Metrics — Single Source of Truth
 
+Machine-readable snapshot: `outputs/metrics_snapshot.json` regenerated with `make metrics-snapshot`.
+
 > **Purpose:** One authoritative table of current pipeline metrics. If any doc disagrees
 > with this file, this file wins. Updated whenever benchmark/benchmark config changes.
 >
@@ -178,7 +180,7 @@ Machine-readable: `outputs/wave0_5_external_predict_results.csv`, `outputs/wave0
 |--------|:-----------------------:|:-------------------:|
 | Ensemble AUROC | 0.7832 | 0.7448 |
 | Ensemble CI₉₅ | 0.717–0.8423 | 0.6741–0.8118 |
-| Expert composite AUROC | 0.7360 | 0.7360 |
+| Expert composite AUROC | 0.7119 | 0.7119 |
 | Expert CI₉₅ | 0.6604–0.8037 | 0.6604–0.8037 |
 | **Delta (expert − ensemble)** | **−0.0472** | **−0.0088** |
 | Verdict | Expert LOWER | Within ±0.02 |
@@ -471,7 +473,7 @@ The MODERATE class (HC50 25-100, n=68) is excluded from the binary task.
 | 2026-06-29 | Wave 0.5b: 23-candidate safety-optimized shortlist (SEED-020–024, no aromatics) | OpenAMP Wave 0.5b |
 | 2026-06-29 | External predictor results filled from wave05_combined_consensus.csv; all 7 gates PASS | OpenAMP Wave 0.5 |
 | 2026-06-29 | Wave 0.5 scaffold diversification — 24-candidate Wave 1 panel across 14 families | OpenAMP Wave 0.5 |
-| 2026-07-01 | Expert ablation benchmark added: expert composite AUROC 0.736 vs ensemble 0.7832 (delta −0.0472); 3 components anti-signal; ensemble remains primary gate | OpenAMP loop |
+| 2026-07-01 | Expert ablation benchmark added: expert composite AUROC 0.7119 vs ensemble 0.7832 (delta −0.0713); anti-signal components documented; ensemble remains primary gate | OpenAMP loop |
 | 2026-07-01 | **Hemolysis benchmark expanded:** 42 -> 238 peptides using DBAASP human erythrocyte data (54 hemolytic vs 125 selective, n=179 binary). Hemolysis risk scorer detection AUROC drops 0.9218 -> 0.5650 (CI 0.47-0.66) — original performance was small-sample inflation. Direction correct, not significant. Safety scorer detection improves 0.3844 -> 0.5116 (still not significant). 196 new peptides from DBAASP v3. | OpenAMP loop |
 | 2026-07-01 | Dedicated hemolysis risk scorer: 4-component score (synth+aromatic+face+cys) achieves detection AUROC=0.9218 (CI: 0.82-0.99); integrated into expert composite (detection 0.5119→0.6429); safety scorer unchanged; 1471 tests | OpenAMP loop |
 | 2026-07-01 | Within-AMP selectivity benchmark added: safety scorer FAILS hemolysis detection (AUROC=0.3844); synthesis is only significant risk detector (AUROC=0.8027); expert composite better than ensemble but not significant (0.5119 vs 0.3486) | OpenAMP loop |

diff --git a/docs/ROADMAP.md b/docs/ROADMAP.md
@@ -141,7 +141,7 @@ Implemented during the pre-wet-lab improvement loop (PRs #31–#54):
 - `run_expert_ablation_benchmark()` in `benchmark/retrospective.py` — scores all 95 AMPs + 96 decoys with both the simple ensemble and the 7-component expert composite
 - `bench expert-ablation` CLI command + `make bench-expert-ablation` target
 - Per-component AUROC attribution: identifies signal-bearing (activity 0.814, selectivity 0.773), near-zero (hinge, novelty, boman, motif), and anti-signal (safety 0.349, serum_stability 0.223, synthesis 0.423) components
-- **Key finding:** Expert composite AUROC 0.736 < ensemble AUROC 0.783 (delta −0.047). The added complexity does NOT improve AMP-vs-decoy discrimination. Three components are anti-signal because real AMPs have extreme biophysical properties that safety/stability/synthesis scorers penalise.
+- **Key finding:** Expert composite AUROC 0.7119 < ensemble AUROC 0.7832 (delta −0.0713). The added complexity does NOT improve AMP-vs-decoy discrimination. Several added components are anti-signal because real AMPs have extreme biophysical properties that safety, stability, and synthesis scorers penalise.
 - Ensemble remains primary synthesis gate. Expert composite may add value for within-AMP ranking (tested in v0.5.9 — see selectivity benchmark; expert composite detection AUROC 0.5119 vs ensemble 0.3486, but not significant at n=14 vs n=21).
 - `make bench-cluster-split` Makefile target added (was missing despite CLI command existing)
 - 14 new tests in `test_expert_ablation.py`; 1435 total tests

diff --git a/src/openamp_foundry/benchmark/metrics_snapshot.py b/src/openamp_foundry/benchmark/metrics_snapshot.py
@@ -0,0 +1,153 @@
+"""Build a machine-readable snapshot of the repo's current benchmark truth."""
+
+from __future__ import annotations
+
+from pathlib import Path
+
+from openamp_foundry.benchmark.retrospective import (
+    run_cluster_split_benchmark,
+    run_expert_ablation_benchmark,
+    run_retrospective_benchmark,
+    run_selectivity_benchmark,
+)
+from openamp_foundry.benchmark.triage import run_triage_benchmark
+
+
+def build_metrics_snapshot(
+    *,
+    amp_csv: str | Path = "examples/validation/known_amps.csv",
+    decoy_csv: str | Path = "examples/validation/random_background.csv",
+    hemolysis_csv: str | Path = "examples/validation/hemolysis_reference.csv",
+    standard_config: str | Path = "configs/pipeline.yaml",
+    phase3_config: str | Path = "configs/phase3.yaml",
+    cluster_threshold: float = 0.70,
+    n_bootstrap: int = 2000,
+) -> dict[str, object]:
+    """Return a compact, auditable snapshot of current benchmark metrics."""
+
+    standard = run_retrospective_benchmark(
+        amp_csv=amp_csv,
+        decoy_csv=decoy_csv,
+        config_path=standard_config,
+        benchmark_type="standard",
+    )
+    phase3 = run_retrospective_benchmark(
+        amp_csv=amp_csv,
+        decoy_csv=decoy_csv,
+        config_path=phase3_config,
+        benchmark_type="standard",
+    )
+    cluster_split = run_cluster_split_benchmark(
+        amp_csv=amp_csv,
+        decoy_csv=decoy_csv,
+        config_path=standard_config,
+        similarity_threshold=cluster_threshold,
+        n_bootstrap=n_bootstrap,
+    )
+    expert_ablation = run_expert_ablation_benchmark(
+        amp_csv=amp_csv,
+        decoy_csv=decoy_csv,
+        config_path=standard_config,
+        n_bootstrap=n_bootstrap,
+    )
+    selectivity = run_selectivity_benchmark(
+        hemolysis_csv=hemolysis_csv,
+        config_path=standard_config,
+        n_bootstrap=n_bootstrap,
+    )
+    triage = run_triage_benchmark(
+        hemolysis_csv=hemolysis_csv,
+        decoy_csv=decoy_csv,
+        config_path=standard_config,
+        n_bootstrap=n_bootstrap,
+    )
+
+    return {
+        "schema_version": 1,
+        "benchmark_inputs": {
+            "amp_csv": str(amp_csv),
+            "decoy_csv": str(decoy_csv),
+            "hemolysis_csv": str(hemolysis_csv),
+            "standard_config": str(standard_config),
+            "phase3_config": str(phase3_config),
+            "cluster_threshold": cluster_threshold,
+            "n_bootstrap": n_bootstrap,
+        },
+        "standard": {
+            "auroc": standard["auroc"],
+            "auprc": standard.get("auprc"),
+            "recall_at_10": standard.get("recall_at_10"),
+            "recall_at_20": standard.get("recall_at_20"),
+            "recall_at_43": standard.get("recall_at_43"),
+            "n_positives": standard["n_positives"],
+            "n_negatives": standard["n_negatives"],
+            "n_total": standard["n_positives"] + standard["n_negatives"],
+            "interpretation": standard["interpretation"],
+        },
+        "phase3": {
+            "auroc": phase3["auroc"],
+            "auprc": phase3.get("auprc"),
+            "n_positives": phase3["n_positives"],
+            "n_negatives": phase3["n_negatives"],
+            "n_total": phase3["n_positives"] + phase3["n_negatives"],
+            "interpretation": phase3["interpretation"],
+        },
+        "cluster_split": {
+            "full_auroc": cluster_split["full_auroc"],
+            "standard_ci95": [
+                cluster_split["standard_ci95_lo"],
+                cluster_split["standard_ci95_hi"],
+            ],
+            "cluster_aware_ci95": [
+                cluster_split["cluster_aware_ci95_lo"],
+                cluster_split["cluster_aware_ci95_hi"],
+            ],
+            "representative_auroc": cluster_split["representative_auroc"],
+            "representative_ci95": [
+                cluster_split["representative_ci95_lo"],
+                cluster_split["representative_ci95_hi"],
+            ],
+            "held_out_auroc": cluster_split["held_out_auroc"],
+            "n_clusters": cluster_split["n_clusters"],
+            "n_multi_member_clusters": cluster_split["n_multi_member_clusters"],
+            "n_amps_in_multi_member_clusters": cluster_split[
+                "n_amps_in_multi_member_clusters"
+            ],
+            "interpretation": cluster_split["interpretation"],
+        },
+        "expert_ablation": {
+            "ensemble_auroc": expert_ablation["ensemble_auroc"],
+            "expert_auroc": expert_ablation["expert_auroc"],
+            "delta_auroc": expert_ablation["delta_auroc"],
+            "signal_bearing_components": expert_ablation[
+                "signal_bearing_components"
+            ],
+            "near_zero_components": expert_ablation["near_zero_components"],
+            "anti_signal_components": expert_ablation["anti_signal_components"],
+            "verdict": expert_ablation["verdict"],
+        },
+        "selectivity": {
+            "n_hemolytic": selectivity["n_hemolytic"],
+            "n_selective": selectivity["n_selective"],
+            "n_border": selectivity["n_border"],
+            "safety_verdict": selectivity["safety_verdict"],
+            "selectivity_proxy_verdict": selectivity["selectivity_proxy_verdict"],
+            "expert_composite_verdict": selectivity["expert_composite_verdict"],
+            "risk_detectors": selectivity["risk_detectors"],
+            "risk_indicators": selectivity["risk_indicators"],
+        },
+        "triage": {
+            "best_scorer": triage["best_scorer"],
+            "top_20_by_ensemble": triage["top_20_by_ensemble"],
+            "top_20_by_triage_score": triage["top_20_by_triage_score"],
+            "per_scorer": {
+                scorer: {
+                    "selective_vs_decoy": info["selective_vs_decoy"]["auroc"],
+                    "hemolytic_vs_decoy": info["hemolytic_vs_decoy"]["auroc"],
+                    "selective_vs_hemolytic": info["selective_vs_hemolytic"]["auroc"],
+                    "triages_correctly": info["triages_correctly"],
+                }
+                for scorer, info in triage["per_scorer"].items()
+            },
+        },
+    }
diff --git a/src/openamp_foundry/cli/commands/benchmark.py b/src/openamp_foundry/cli/commands/benchmark.py
@@ -219,3 +219,24 @@ def _run_triage(args: argparse.Namespace) -> int:
     }
     print(json.dumps(summary, indent=2))
     return 0
+
+
+def _run_metrics_snapshot(args: argparse.Namespace) -> int:
+    import json
+
+    from openamp_foundry.benchmark.metrics_snapshot import build_metrics_snapshot
+    from openamp_foundry.utils.io import write_json
+
+    result = build_metrics_snapshot(
+        amp_csv=args.amp_csv,
+        decoy_csv=args.decoy_csv,
+        hemolysis_csv=args.hemolysis_csv,
+        standard_config=args.config,
+        phase3_config=args.phase3_config,
+        cluster_threshold=args.threshold,
+        n_bootstrap=args.n_bootstrap,
+    )
+    if args.out:
+        write_json(args.out, result)
+    print(json.dumps(result, indent=2))
+    return 0
diff --git a/src/openamp_foundry/cli/main.py b/src/openamp_foundry/cli/main.py
@@ -1,7 +1,7 @@
 from __future__ import annotations
 
 from openamp_foundry.cli.commands.core import _run_generate_batch
-from openamp_foundry.cli.commands.benchmark import _run_bench, _run_validate_scoring, _run_cluster_split_bench, _run_expert_ablation_bench, _run_selectivity_bench, _run_triage
+from openamp_foundry.cli.commands.benchmark import _run_bench, _run_validate_scoring, _run_cluster_split_bench, _run_expert_ablation_bench, _run_selectivity_bench, _run_triage, _run_metrics_snapshot
 from openamp_foundry.cli.commands.selection import _run_pilot_panel, _run_pilot_confident, _run_diversity_check
 from openamp_foundry.cli.commands.external import _run_external_predict, _run_external_consensus
 from openamp_foundry.cli.commands.qc import _run_synthesis_order, _run_presynth_qc
@@ -207,6 +207,35 @@ def build_parser() -> argparse.ArgumentParser:
         help="Optional JSON output path.",
     )
 
+    metrics_snapshot = bench_sub.add_parser(
+        "metrics-snapshot",
+        help=(
+            "Build a machine-readable snapshot of the repo's current benchmark truth. "
+            "Use this to refresh docs/METRICS_CURRENT.md evidence and catch number drift."
+        ),
+    )
+    metrics_snapshot.add_argument(
+        "--amp-csv",
+        default="examples/validation/known_amps.csv",
+    )
+    metrics_snapshot.add_argument(
+        "--decoy-csv",
+        default="examples/validation/random_background.csv",
+    )
+    metrics_snapshot.add_argument(
+        "--hemolysis-csv",
+        default="examples/validation/hemolysis_reference.csv",
+    )
+    metrics_snapshot.add_argument("--config", default="configs/pipeline.yaml")
+    metrics_snapshot.add_argument("--phase3-config", default="configs/phase3.yaml")
+    metrics_snapshot.add_argument("--threshold", type=float, default=0.70)
+    metrics_snapshot.add_argument("--n-bootstrap", type=int, default=2000)
+    metrics_snapshot.add_argument(
+        "--out",
+        required=False,
+        help="Optional JSON output path.",
+    )
+
     generate = sub.add_parser(
         "generate-batch",
         help=(
@@ -659,6 +688,8 @@ def main(argv: list[str] | None = None) -> int:
             return _run_selectivity_bench(args)
         if args.bench_command == "triage":
             return _run_triage(args)
+        if args.bench_command == "metrics-snapshot":
+            return _run_metrics_snapshot(args)
         return _run_bench(args)
 
     if args.command == "generate-batch":
@@ -713,4 +744,3 @@ def main(argv: list[str] | None = None) -> int:
 if __name__ == "__main__":
     raise SystemExit(main())
 
-
diff --git a/tests/test_docs_consistent.py b/tests/test_docs_consistent.py
@@ -3,6 +3,7 @@
 from __future__ import annotations
 
 import re
+import json
 from pathlib import Path
 
 
@@ -17,6 +18,7 @@
 CURRENT_N_TOTAL = 191
 EXPANDED_PR_REF = "PR #110"
 ORIGINAL_DEMO_AUROC = "0.8420"
+SNAPSHOT_PATH = REPO_ROOT / "outputs" / "metrics_snapshot.json"
 
 # Patterns that indicate stale "current" usage (not historical context)
 _STALE_CURRENT_PATTERNS: list[str] = [
@@ -91,6 +93,16 @@ def test_doc_metrics_current_md_is_authoritative(self):
             f"METRICS_CURRENT.md missing n={CURRENT_N_TOTAL}"
         )
 
+    def test_doc_metrics_snapshot_matches_authoritative_values(self):
+        """Machine-readable metrics snapshot must agree with docs' current truth."""
+        assert SNAPSHOT_PATH.exists(), "outputs/metrics_snapshot.json not found"
+        payload = json.loads(SNAPSHOT_PATH.read_text(encoding="utf-8"))
+        assert payload["standard"]["auroc"] == float(CURRENT_AUROC)
+        assert payload["phase3"]["auroc"] == float(CURRENT_PHASE3_AUROC)
+        assert payload["standard"]["n_positives"] == CURRENT_N_POSITIVES
+        assert payload["standard"]["n_negatives"] == CURRENT_N_NEGATIVES
+        assert payload["standard"]["n_total"] == CURRENT_N_TOTAL
+
     def test_doc_no_breaking_news_terminology(self):
         """'breaking news' must be replaced with 'high-impact scenario' in all docs."""
         breaking = re.compile(r"breaking[-\s]news", re.IGNORECASE)

diff --git a/tests/test_metrics_snapshot.py b/tests/test_metrics_snapshot.py
@@ -0,0 +1,25 @@
+from __future__ import annotations
+
+import json
+
+from openamp_foundry.benchmark.metrics_snapshot import build_metrics_snapshot
+
+
+def test_metrics_snapshot_matches_current_benchmark_truth(tmp_path):
+    snapshot = build_metrics_snapshot(n_bootstrap=200)
+
+    assert snapshot["schema_version"] == 1
+    assert snapshot["standard"]["n_total"] == 191
+    assert snapshot["standard"]["n_positives"] == 95
+    assert snapshot["standard"]["n_negatives"] == 96
+    assert snapshot["standard"]["auroc"] == 0.7832
+    assert snapshot["phase3"]["auroc"] == 0.7448
+    assert snapshot["cluster_split"]["full_auroc"] == 0.7832
+    assert snapshot["expert_ablation"]["expert_auroc"] == 0.7119
+    assert snapshot["selectivity"]["n_hemolytic"] == 54
+    assert snapshot["triage"]["best_scorer"] == "selectivity_proxy"
+
+    out = tmp_path / "metrics_snapshot.json"
+    out.write_text(json.dumps(snapshot, indent=2, sort_keys=True), encoding="utf-8")
+    reloaded = json.loads(out.read_text(encoding="utf-8"))
+    assert reloaded["standard"]["auroc"] == 0.7832