sync EmbedEd artifacts (phase01-v8)

artifacts/sprime_gate_runs.json

@@ -92,6 +92,55 @@
           "n_anchors": 800
         }
       }
+    },
+    {
+      "stamp": "2026-10-03T08:01:15Z",
+      "version": "phase01-v8",
+      "rule": "D1 scale-aware: C1 < C2 <= C3 and d(C1->C2) >= 0.5",
+      "ok": true,
+      "d_c1_c2": 0.8863,
+      "ci": [
+        1.337,
+        1.643
+      ],
+      "per_condition": {
+        "C1": {
+          "mean_cos": 0.93383,
+          "std": 0.03924,
+          "n_samples": 16000,
+          "n_anchors": 800
+        },
+        "C2": {
+          "mean_cos": 0.96355,
+          "std": 0.02663,
+          "n_samples": 16000,
+          "n_anchors": 800
+        },
+        "C3": {
+          "mean_cos": 0.9808,
+          "std": 0.01139,
+          "n_samples": 16000,
+          "n_anchors": 800
+        },
+        "C4": {
+          "mean_cos": 0.95653,
+          "std": 0.02799,
+          "n_samples": 16000,
+          "n_anchors": 800
+        },
+        "C5": {
+          "mean_cos": 0.93355,
+          "std": 0.03927,
+          "n_samples": 16000,
+          "n_anchors": 800
+        },
+        "C6": {
+          "mean_cos": 0.97507,
+          "std": 0.01279,
+          "n_samples": 16000,
+          "n_anchors": 800
+        }
+      }
     }
   ]
 }

report/measurements.md

@@ -648,7 +648,7 @@ Generated from `artifacts/runs/<condition>_<seed>/eval_metrics.json` by `python
 
 ## s′ census — measured pair counts and the holdout (D4)
 
-Measured from `data_bcb_v2_sampled_bf.pickle` (md5 `c5495307a422…`, 4600 rows) by `python -m embeded.generalize --census --holdout-k 3` — version `phase01-v6`, recorded before any F1 exists (PHASE3_PLAN §3.2, decision D4).
+Measured from `data_bcb_v2_sampled_bf.pickle` (md5 `c5495307a422…`, 4600 rows) by `python -m embeded.generalize --census --holdout-k 3` — version `phase01-v8`, recorded before any F1 exists (PHASE3_PLAN §3.2, decision D4).
 
 | functionality | pairs | clone | non-clone |
 |---|---|---|---|
@@ -680,11 +680,14 @@ Measured from `data_bcb_v2_sampled_bf.pickle` (md5 `c5495307a422…`, 4600 rows)
 
 ## Gate G1s — hardness on s′
 
-### run 2026-09-28T15:32:45Z — version `phase01-v6`, rule: D1 scale-aware: C1 < C2 <= C3 and d(C1->C2) >= 0.5
+### run 2026-10-03T08:01:15Z — version `phase01-v8`, rule: D1 scale-aware: C1 < C2 <= C3 and d(C1->C2) >= 0.5
 
 - C1: mean cos(anchor, negative) = 0.93383 (sd 0.039, n = 16000, anchors = 800)
 - C2: mean cos(anchor, negative) = 0.96355 (sd 0.027, n = 16000, anchors = 800)
 - C3: mean cos(anchor, negative) = 0.98080 (sd 0.011, n = 16000, anchors = 800)
+- C4: mean cos(anchor, negative) = 0.95653 (sd 0.028, n = 16000, anchors = 800)
+- C5: mean cos(anchor, negative) = 0.93355 (sd 0.039, n = 16000, anchors = 800)
+- C6: mean cos(anchor, negative) = 0.97507 (sd 0.013, n = 16000, anchors = 800)
 - embeddings: microsoft/graphcodebert-base max_len=512 n=4140
 - corpus: s′ (bcb_v2_sampled_bf) — the CodeXGLUE verdict must NOT be assumed to transfer; this gate re-measures it (PHASE3_PLAN §2 item 4)
 - 95% CI for d(C1→C2) = 1.337–1.643 (2000 resamples over anchors, not negatives)

sync_manifest.json

@@ -1,6 +1,6 @@
 {
   "version": "phase01-v8",
-  "synced_at_utc": "2026-10-03T07:56:23Z",
+  "synced_at_utc": "2026-10-03T08:04:33Z",
   "staged_files": [
     "artifacts/audit/audit_key.csv",
     "artifacts/audit/audit_labels.csv",
@@ -344,7 +344,8 @@
     "artifacts/v5-backup/mining_summary.json",
     "artifacts/v5-backup/triples_C1.jsonl",
     "artifacts/v5-backup/triples_C2.jsonl",
-    "artifacts/v5-backup/triples_C3.jsonl"
+    "artifacts/v5-backup/triples_C3.jsonl",
+    "report/measurements.md"
   ],
   "excluded_patterns": [
     "sprime/*",