Metadata: size_categories, provenance replay wording, contributor name, collection

README.md

@@ -12,6 +12,8 @@ tags:
 - curation
 - provenance
 - research-snapshot
+size_categories:
+- n<1K
 configs:
 - config_name: viewer
   data_files:

provenance.json

@@ -1,17 +1,17 @@
 {
   "schema_version": "1.0.0",
   "dataset_id": "rmems/python-function-repair-20261006",
-  "purpose": "Deterministic Python function-repair records: a mutator breaks real MIT-licensed programs, the harness executes original vs mutant, the repair is verified by replay",
+  "purpose": "Deterministic Python function-repair records: a mutator breaks real MIT-licensed programs, the harness executes original vs mutant, the repair is checked by replay (partial coverage, see raw_snapshot)",
   "primary_target": "Research retention, evaluation, and degradation-vs-supervision comparison; curation pending before any training use",
   "generation": {
     "primary_model": null,
     "generator": "python-repair-mutator 2.0.0 (project-owned)",
     "contribution": "Generated every record in this release",
-    "method": "pipelines/code_repair_cli.py generate, catalog python-repair-v1, seed 20261006, count 150; replay passed"
+    "method": "pipelines/code_repair_cli.py generate, catalog python-repair-v1, seed 20261006, count 150; replay covered 97 of 132 candidates (97 passed, 0 failed, 35 not replayed)"
   },
   "contributors": [
     {
-      "name": "Raul Montoya Cardenas",
+      "name": "Raul Cardenas Montoya",
       "roles": [
         "operator",
         "review"
@@ -27,7 +27,7 @@
   "source_factory": "python-function-repair-factory",
   "source_repository": "https://github.com/rmems/synthetic-factory",
   "source_run": "outputs/code-repair/cr-20261006/",
-  "collection": null,
+  "collection": "rmems/synthetic-data-factory-procedural-and-deterministic-6ac6d580a7f8dd780d70505e",
   "payload_published": true,
   "training_ready": false,
   "raw_evidence_policy": "Public raw JSONL evidence is published under data/raw; harness telemetry lives under evidence/ and is not viewer rows; curated training export remains pending.",
@@ -46,7 +46,7 @@
   "raw_snapshot": {
     "run": "cr-20261006",
     "generator": "python-repair-mutator",
-    "selection_rule": "candidates.jsonl verbatim (132 of 150 requested; 18 duplicate mutants + 4 siteless programs skipped). execution-log.jsonl is harness telemetry, excluded from viewer rows. Pre-replay candidates are the minimally-curated view; replay passed.",
+    "selection_rule": "candidates.jsonl verbatim (132 candidates, 110 accepted and 22 rejected, from 150 requested draws; 18 draws skipped as duplicate mutants; separately, 4 of the 197 catalog programs have no mutation sites and were excluded from the draw pool). Oracle statuses: 115 validated, 17 provisional. execution-log.jsonl is harness telemetry, excluded from viewer rows. Pre-replay candidates are the minimally-curated view; replay covered 97 of 132 (35 not replayed: all 22 rejected rows plus 13 accepted rows whose hidden check was unavailable).",
     "omitted_not_published": [
       "execution-log.jsonl from viewer rows (kept under evidence/)"
     ],
@@ -68,7 +68,7 @@
     }
   },
   "gate_outcomes": {
-    "replay": "passed",
+    "replay": "passed for replayed subset (97 of 132 replayed, 0 failed, 35 not replayed)",
     "validate_run": "0 errors (candidates.jsonl)",
     "check_records": "0 errors, 0 warnings (candidates.jsonl)"
   }