Record immutable benchmark revision in run manifests

README.md

@@ -91,6 +91,7 @@ Use the first-class quant gate for every candidate before spending hours on the
   --model '<served-model-id>' \
   --hf-repo '<owner/public-model-repo>' \
   --hf-revision '<full-published-commit-sha>' \
+  --benchmark-revision '<full-shiftedx-bench-commit-sha>' \
   --profile vision \
   --label '<stable-candidate-label>' \
   --output-dir 'results/<stable-candidate-label>'

src/shiftedx_bench/cli.py

@@ -78,10 +78,17 @@ def _reasoning_request_overrides(args: argparse.Namespace) -> dict[str, Any]:
 def _record_evaluated_artifact(manifest: dict[str, Any], args: argparse.Namespace) -> None:
     repo_id = getattr(args, "hf_repo", None)
     revision = getattr(args, "hf_revision", None)
+    benchmark_revision = getattr(args, "benchmark_revision", None)
     if bool(repo_id) != bool(revision):
         raise ValueError("--hf-repo and --hf-revision must be supplied together")
+    if repo_id and not benchmark_revision:
+        raise ValueError("--benchmark-revision is required for a post-publication run")
     if repo_id:
         manifest["evaluated_artifact"] = {"repo_id": repo_id, "revision": revision}
+        manifest["benchmark_source"] = {
+            "repo_id": "Shiftedx/shiftedx-bench",
+            "revision": benchmark_revision,
+        }
 
 
 def _validate_context_variant_profile(args: argparse.Namespace, variants: list[dict[str, Any]]) -> None:
@@ -229,6 +236,7 @@ def run_quant_gate(args: argparse.Namespace) -> None:
         "model": args.model,
         "hf_repo": args.hf_repo,
         "hf_revision": args.hf_revision,
+        "benchmark_revision": args.benchmark_revision,
         "api_key_env": args.api_key_env,
         "timeout": args.timeout,
         "variants": None,
@@ -280,6 +288,7 @@ def render_model_card(args: argparse.Namespace) -> None:
         profile=args.profile,
         model_repo=args.model_repo,
         model_revision=args.model_revision,
+        benchmark_revision=args.benchmark_revision,
     )
     block = render_benchmark_block(
         rows,
@@ -398,6 +407,7 @@ def make_parser() -> argparse.ArgumentParser:
     command.add_argument("--profile", choices=["vision", "text"], default="vision")
     command.add_argument("--hf-repo", help="Published Hugging Face repo evaluated by this run")
     command.add_argument("--hf-revision", help="Full immutable Hugging Face revision evaluated by this run")
+    command.add_argument("--benchmark-revision", help="Full immutable Shiftedx Bench revision used by this run")
     command.set_defaults(func=run_quant_gate)
 
     command = subparsers.add_parser("render-model-card")

src/shiftedx_bench/model_card.py

@@ -54,6 +54,7 @@ def validate_post_publish_run(
     profile: str,
     model_repo: str,
     model_revision: str,
+    benchmark_revision: str,
 ) -> tuple[list[dict[str, Any]], str]:
     expected = EXPECTED_COUNTS[profile]
     rows = [row for row in _load_rows(run_dir, expected) if str(row.get("variant")) == variant]
@@ -89,6 +90,12 @@ def validate_post_publish_run(
                 f"Post-publication provenance mismatch in {path.name}: "
                 f"expected {model_repo}@{model_revision}, found {artifact}"
             )
+        benchmark = manifest.get("benchmark_source") or {}
+        if benchmark.get("repo_id") != "Shiftedx/shiftedx-bench" or benchmark.get("revision") != benchmark_revision:
+            raise ValueError(
+                f"Benchmark provenance mismatch in {path.name}: expected "
+                f"Shiftedx/shiftedx-bench@{benchmark_revision}, found {benchmark}"
+            )
     if len(versions) != 1 or "None" in versions:
         raise ValueError(f"Run manifests disagree on benchmark version: {sorted(versions)}")
     return rows, versions.pop()

tests/test_model_card.py

@@ -56,6 +56,10 @@ def _write_run(tmp_path, *, omit_lane=None):
             "benchmark_version": "0.3.0",
             "variants": [{"label": VARIANT, "request_overrides": {}}],
             "evaluated_artifact": {"repo_id": MODEL_REPO, "revision": MODEL_REVISION},
+            "benchmark_source": {
+                "repo_id": "Shiftedx/shiftedx-bench",
+                "revision": BENCHMARK_REVISION,
+            },
         }
         (tmp_path / f"{names[lane]}.manifest.json").write_text(json.dumps(manifest), encoding="utf-8")
 
@@ -68,6 +72,7 @@ def test_post_publish_model_card_block_is_complete_and_idempotent(tmp_path):
         profile="vision",
         model_repo=MODEL_REPO,
         model_revision=MODEL_REVISION,
+        benchmark_revision=BENCHMARK_REVISION,
     )
     block = render_benchmark_block(
         rows,
@@ -101,6 +106,7 @@ def test_model_card_export_rejects_incomplete_or_wrong_revision(tmp_path):
             profile="vision",
             model_repo=MODEL_REPO,
             model_revision=MODEL_REVISION,
+            benchmark_revision=BENCHMARK_REVISION,
         )
     with pytest.raises(ValueError, match="provenance mismatch"):
         validate_post_publish_run(
@@ -109,4 +115,5 @@ def test_model_card_export_rejects_incomplete_or_wrong_revision(tmp_path):
             profile="text",
             model_repo=MODEL_REPO,
             model_revision="c" * 40,
+            benchmark_revision=BENCHMARK_REVISION,
         )