updating purpose-classifier data:

data/generation-manifest.json

@@ -1,10 +1,9 @@
 {
-  "schemaVersion": 1,
-  "dataset": "purpose-classifier-source-v1",
   "canonicalFiles": [
     "purpose-prompts.jsonl",
     "purpose-prompts-round2.jsonl"
   ],
+  "dataset": "purpose-classifier-source-v1",
   "derivedBatchFiles": [
     "round2-01.jsonl",
     "round2-02.jsonl",
@@ -24,34 +23,52 @@
   ],
   "generations": [
     {
+      "date": "2026-07-29",
       "file": "purpose-prompts.jsonl",
       "model": "mixed frontier-model runs (legacy sol/opus aliases; exact model IDs were not retained)",
-      "date": "2026-07-29",
+      "notes": "The canonical aggregate was curated from the original per-model batches. Three exact overlaps with the shipped eval fixture were removed on 2026-07-30.",
       "prompt": "../datagen-prompt.md",
       "topics": [
         "web and mobile",
         "backend and data",
         "infrastructure and systems",
         "developer tooling"
-      ],
-      "notes": "The canonical aggregate was curated from the original per-model batches. Three exact overlaps with the shipped eval fixture were removed on 2026-07-30."
+      ]
     },
     {
+      "date": "2026-07-30",
       "file": "purpose-prompts-round2.jsonl",
       "model": "Nucleic frontier-model generator (exact underlying model ID was not retained)",
-      "date": "2026-07-30",
+      "notes": "Corrective generation that counterbalances round-one label, slice, length, opener, and language drift.",
       "prompt": "../datagen-prompt-2.md",
       "topics": [
         "boundary confusion pairs",
         "pasted context",
         "mixed intent",
         "non-English developer prompts"
-      ],
-      "notes": "Corrective generation that counterbalances round-one label, slice, length, opener, and language drift."
+      ]
     }
   ],
+  "labeling": {
+    "date": "2026-08-02",
+    "files": {
+      "purpose-prompts-round2.jsonl": {
+        "records": 3000,
+        "sha256": "a3c5bbede5b7db743c9f31756491990b388edd2aa9b41c8b2b9ec95b7115f4aa"
+      },
+      "purpose-prompts.jsonl": {
+        "records": 9007,
+        "sha256": "6fbd08df113ca4c6f9da151222772dc3e25e1b9b19c9c2b7df9f572053af2507"
+      }
+    },
+    "model": "gpt-5.6-sol",
+    "reasoningEffort": "high",
+    "schemaVersion": 1,
+    "scope": "all canonical public prompts; rejected junk removed"
+  },
   "limitations": [
     "The exact generator model IDs and sampling parameters were not recorded when the source corpora were created.",
     "The round2-NN files are retained generation batches and duplicate the round-two canonical aggregate; dataset tooling must read canonicalFiles only."
-  ]
+  ],
+  "schemaVersion": 1
 }