Update main checkpoint with focused extraction SFT and verified results

Publish complete verified main weights and native inference package. Add continuation lineage, matched results and four checked examples; retain the existing public model name and release history.

Files changed (16) hide show

CITATION.cff +1 -1
INFERENCE_VERIFICATION.json +2515 -287
JSONLITE_RESULTS.json +0 -0
PACKAGING_VERIFICATION.json +141 -482
README.md +78 -117
RUN_GUIDE.md +5 -2
SHA256SUMS +15 -15
SOURCE_NOTICE.md +9 -5
TECHNICAL_REPORT.md +93 -138
checkpoint_metadata.json +34 -20
checkpoint_provenance.json +17 -5
complete-request-exact.svg +1 -1
config.json +5 -3
inference.py +1 -1
model.safetensors +1 -1
native_bundle.zip +2 -2

CITATION.cff CHANGED Viewed

@@ -4,7 +4,7 @@ title: "OpenSML-JSONLite-175M: Native JSON Generation from Source Text and Schem
 authors:
   - family-names: Zebrowski
     given-names: William
-version: 0.2.0
 date-released: 2026-10-14
 url: "https://github.com/williamzebrowskI/OpenSML-JSONLite-175M"
 repository-code: "https://github.com/williamzebrowskI/OpenSML-JSONLite-175M"

 authors:
   - family-names: Zebrowski
     given-names: William
+version: 0.3.0
 date-released: 2026-10-14
 url: "https://github.com/williamzebrowskI/OpenSML-JSONLite-175M"
 repository-code: "https://github.com/williamzebrowskI/OpenSML-JSONLite-175M"

INFERENCE_VERIFICATION.json CHANGED Viewed

@@ -1,454 +1,2682 @@
 {
-  "constrained_decoding": false,
-  "correct_against_reference": 42,
-  "json_assembly": false,
   "model": "OpenSML-JSONLite-175M-Instruct",
-  "model_calls_per_request": 1,
-  "output_mode": "native-json",
-  "output_repair": false,
   "outputs": [
     {
-      "correct": false,
-      "id": "FIELD-01",
-      "index": 1,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"playlist_owner\":[],\"playlist\":[\"Rock Gaming\"],\"entity_name\":[\"Mariem Hassan\"]}"
     },
     {
-      "correct": false,
-      "id": "FIELD-02",
-      "index": 2,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"music_item\":[],\"artist\":[\"Mariem Hassan\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-03",
-      "index": 3,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"playlist_owner\":[\"my\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-04",
-      "index": 4,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"artist\":[],\"entity_name\":[],\"playlist\":[\"Digster Future Hits\"]}"
     },
     {
-      "correct": false,
-      "id": "FIELD-05",
-      "index": 5,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"state\":[\"AR\"],\"party_size_number\":[\"nine\"],\"state\":[\"AR\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-06",
-      "index": 6,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"party_size_number\":[\"nine\"],\"timeRange\":[],\"sort\":[]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-07",
-      "index": 7,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"sort\":[\"best\"],\"city\":[\"Wrigley\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-08",
-      "index": 8,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"timeRange\":[],\"poi\":[]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-09",
-      "index": 9,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"condition_temperature\":[\"colder\"],\"state\":[\"Louisiana\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-10",
-      "index": 10,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"current_location\":[],\"geographic_poi\":[],\"condition_temperature\":[\"colder\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-11",
-      "index": 11,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"timeRange\":[\"last August\"],\"geographic_poi\":[\"Stunt Ranch Santa Monica Mountains Reserve\"],\"condition_temperature\":[\"colder\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-12",
-      "index": 12,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"current_location\":[],\"geographic_poi\":[\"Stunt Ranch Santa Monica Mountains Reserve\"],\"state\":[]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-13",
-      "index": 13,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"artist\":[\"Behzad Mirkhani\"],\"music_item\":[\"theme\"],\"sort\":[\"last\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-14",
-      "index": 14,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"playlist\":[],\"music_item\":[\"theme\"],\"service\":[]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-15",
-      "index": 15,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"artist\":[\"Yui\"],\"music_item\":[\"track\"],\"service\":[\"Vimeo\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-16",
-      "index": 16,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"sort\":[],\"playlist\":[]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-17",
-      "index": 17,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"rating_value\":[\"3\"],\"rating_unit\":[\"stars\"],\"object_name\":[\"Things Fall Apart\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-18",
-      "index": 18,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"object_name\":[\"Things Fall Apart\"],\"object_select\":[],\"object_type\":[]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-19",
-      "index": 19,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"rating_value\":[\"1\"],\"object_name\":[]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-20",
-      "index": 20,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"object_part_of_series_type\":[],\"object_name\":[]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-21",
-      "index": 21,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"object_name\":[\"Happiness Was Free\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-22",
-      "index": 22,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"object_type\":[],\"object_name\":[\"Happiness Was Free\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-23",
-      "index": 23,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"object_name\":[\"Martin Morning\"],\"object_type\":[\"saga\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-24",
-      "index": 24,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"object_name\":[\"Martin Morning\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-25",
-      "index": 25,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"movie_type\":[\"animated movies\"],\"spatial_relation\":[\"in the neighborhood\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-26",
-      "index": 26,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"movie_name\":[],\"timeRange\":[],\"movie_type\":[\"animated movies\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-27",
-      "index": 27,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"timeRange\":[\"twenty two o'clock\"],\"movie_name\":[\"Mabel’s Dramatic Career\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-28",
-      "index": 28,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"timeRange\":[\"twenty two o'clock\"],\"spatial_relation\":[],\"object_location_type\":[]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-29",
-      "index": 29,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"album\":[],\"artist\":[\"Mira Sol\"],\"track\":[\"Blue Lantern\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-30",
-      "index": 30,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"restaurant_name\":[\"River Moon\"],\"city\":[\"Denver\"],\"party_size_number\":[\"four\"],\"timeRange\":[\"tomorrow\"]}"
     },
     {
-      "correct": false,
-      "id": "FIELD-31",
-      "index": 31,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"country\":[],\"timeRange\":[\"Friday\"],\"city\":[\"Odense\"]}"
     },
     {
-      "correct": false,
-      "id": "FIELD-32",
-      "index": 32,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"object_name\":[\"Glass Harbor\"],\"rating_value\":[\"four\"],\"best_rating\":[],\"rating_unit\":[\"stars\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-33",
-      "index": 33,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"timeRange\":[],\"location_name\":[\"Maple Cinema\"],\"movie_name\":[\"Silver Orchard\"]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-34",
-      "index": 34,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"entity_name\":[\"Night Steps\"],\"playlist\":[\"Evening Walks\"]}"
     },
     {
-      "correct": false,
-      "id": "FIELD-35",
-      "index": 35,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"artist\":[\"Mira Sol and Theo Lane\"],\"album\":[]}"
     },
     {
-      "correct": false,
-      "id": "FIELD-36",
-      "index": 36,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"city\":[\"Bergen\"],\"country\":[]}"
     },
     {
-      "correct": false,
-      "id": "FIELD-37",
-      "index": 37,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"cuisine\":[],\"city\":[\"Boston\"],\"restaurant_name\":[]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-38",
-      "index": 38,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"album\":[\"Quiet Signals\"],\"artist\":[\"Mira Sol\"],\"track\":[]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-39",
-      "index": 39,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"object_name\":[\"Paper Bridge\"],\"rating_value\":[]}"
     },
     {
-      "correct": true,
-      "id": "FIELD-40",
-      "index": 40,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"country\":[],\"timeRange\":[],\"artist\":[]}"
     },
     {
-      "correct": true,
-      "id": "native:snips:AddToPlaylist:train:80:view1:native:alias",
-      "index": 41,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"selected_playlist\":[\"night vibes\"],\"selected_artist\":[],\"selected_playlist_owner\":[]}"
     },
     {
-      "correct": false,
-      "id": "native:snips:AddToPlaylist:train:80:view1:native:nullable",
-      "index": 42,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"playlist_owner\":null,\"artist\":\"gn\",\"playlist\":night vibes\"}"
     },
     {
-      "correct": false,
-      "id": "native:snips:BookRestaurant:train:684:view1:native:alias",
-      "index": 43,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"selected_restaurant_name\":[\"mughlai restaurant\"],\"selected_timeRange\":[\"Nov. the 1st\"],\"selected_sort\":[]}"
     },
     {
-      "correct": false,
-      "id": "native:snips:BookRestaurant:train:684:view1:native:nullable",
-      "index": 44,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"restaurant_name\":\"Mughlai Restaurant\",\"timeRange\":\"Nov. the 1st\",\"sort\":null}"
     },
     {
-      "correct": true,
-      "id": "native:snips:GetWeather:train:109:view0:native:alias",
-      "index": 45,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"selected_condition_temperature\":[\"colder\"],\"selected_state\":[\"Louisiana\"]}"
     },
     {
-      "correct": true,
-      "id": "native:snips:GetWeather:train:109:view0:native:nullable",
-      "index": 46,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"state\":\"Louisiana\",\"condition_temperature\":\"colder\"}"
     },
     {
-      "correct": true,
-      "id": "native:snips:PlayMusic:train:801:view1:native:alias",
-      "index": 47,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"selected_playlist\":[],\"selected_music_item\":[\"theme\"],\"selected_service\":[]}"
     },
     {
-      "correct": true,
-      "id": "native:snips:PlayMusic:train:801:view1:native:nullable",
-      "index": 48,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"playlist\":null,\"music_item\":\"theme\",\"service\":null}"
     },
     {
-      "correct": true,
-      "id": "native:snips:RateBook:train:293:view0:native:alias",
-      "index": 49,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"selected_rating_value\":[\"2\"],\"selected_object_select\":[\"current\"]}"
     },
     {
-      "correct": true,
-      "id": "native:snips:RateBook:train:293:view0:native:nullable",
-      "index": 50,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"object_select\":\"current\",\"rating_value\":\"2\"}"
     },
     {
-      "correct": true,
-      "id": "native:snips:SearchCreativeWork:train:602:view1:native:alias",
-      "index": 51,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"selected_object_name\":[\"Playlist: The Very Best of Dolly Parton\"]}"
     },
     {
-      "correct": true,
-      "id": "native:snips:SearchCreativeWork:train:602:view1:native:nullable",
-      "index": 52,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"object_name\":\"Playlist: The Very Best of Dolly Parton\"}"
     },
     {
-      "correct": true,
-      "id": "native:snips:SearchScreeningEvent:train:897:view1:native:alias",
-      "index": 53,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"selected_movie_type\":[\"animated movies\"],\"selected_movie_name\":[],\"selected_location_name\":[]}"
     },
     {
-      "correct": false,
-      "id": "native:snips:SearchScreeningEvent:train:897:view1:native:nullable",
-      "index": 54,
-      "raw_parity": true,
       "stop_reason": "turn_end",
-      "text": "{\"movie_name\":null,\"movie_type\":\"movies\",\"location_name\":null}"
     }
   ],
-  "raw_text_and_stop_matches": 54,
-  "requests": 54,
-  "scope": "Clean stock-MLX packaging parity on observed playground fixtures; not full-heldout runtime parity or fresh blind accuracy.",
-  "step": 469,
-  "versions": {
-    "jsonschema": "4.26.0",
-    "mlx": "0.32.3",
-    "numpy": "2.4.2",
-    "tokenizers": "0.22.2"
-  },
-  "weight_sha256": "656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4"
 }

 {
   "model": "OpenSML-JSONLite-175M-Instruct",
+  "checkpoint": "main",
+  "weight_sha256": "655fb5cfba009163cd8a7f7fe7b170b998bb452c217b1b473cfe10b33de0ccd1",
+  "requests": 54,
+  "raw_text_and_stop_matches": 54,
+  "correct_against_reference": 44,
+  "schema_valid": 54,
+  "versions": {
+    "mlx": "0.32.3",
+    "numpy": "2.4.2",
+    "tokenizers": "0.22.2",
+    "jsonschema": "4.26.0"
+  },
+  "platform": "Apple Silicon macOS, Python3.11",
   "outputs": [
     {
+      "text": "{\"playlist_owner\":[],\"playlist\":[\"Rock Gaming\"],\"entity_name\":[\"Mariem Hassan\"]}",
+      "value": {
+        "playlist_owner": [],
+        "playlist": [
+          "Rock Gaming"
+        ],
+        "entity_name": [
+          "Mariem Hassan"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 37,
+      "seconds": 0.1317074578255415,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-01",
+      "checks": {
+        "exact": 0,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 1,
+        "fields": 3,
+        "tp": 1,
+        "fp": 1,
+        "fn": 2,
+        "present_correct": 1,
+        "present_total": 3,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 1,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"artist\":[\"Mariem Hassan\"],\"music_item\":[]}",
+      "value": {
+        "artist": [
+          "Mariem Hassan"
+        ],
+        "music_item": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 23,
+      "seconds": 0.09170329081825912,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-02",
+      "checks": {
+        "exact": 0,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 1,
+        "fields": 2,
+        "tp": 0,
+        "fp": 1,
+        "fn": 0,
+        "present_correct": 0,
+        "present_total": 0,
+        "absent_correct": 1,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 1,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"playlist_owner\":[\"my\"]}",
+      "value": {
+        "playlist_owner": [
+          "my"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 13,
+      "seconds": 0.052787666907534,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-03",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 1,
+        "fields": 1,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"artist\":[],\"entity_name\":[],\"playlist\":[\"Digster Future Hits\"]}",
+      "value": {
+        "artist": [],
+        "entity_name": [],
+        "playlist": [
+          "Digster Future Hits"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 29,
+      "seconds": 0.09026900003664196,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-04",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"state\":[\"AR\"],\"party_size_number\":[\"nine\"],\"restaurant_type\":[\"brasserie\"]}",
+      "value": {
+        "state": [
+          "AR"
+        ],
+        "party_size_number": [
+          "nine"
+        ],
+        "restaurant_type": [
+          "brasserie"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 36,
+      "seconds": 0.13286249991506338,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-05",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 3,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 3,
+        "present_total": 3,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 3,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"party_size_number\":[\"nine\"],\"timeRange\":[],\"sort\":[]}",
+      "value": {
+        "party_size_number": [
+          "nine"
+        ],
+        "timeRange": [],
+        "sort": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 27,
+      "seconds": 0.11090816603973508,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-06",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"city\":[\"Wrigley\"],\"sort\":[\"best\"]}",
+      "value": {
+        "city": [
+          "Wrigley"
+        ],
+        "sort": [
+          "best"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 21,
+      "seconds": 0.08733266708441079,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-07",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 2,
+        "present_total": 2,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"poi\":[],\"timeRange\":[]}",
+      "value": {
+        "poi": [],
+        "timeRange": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 15,
+      "seconds": 0.06839366699568927,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-08",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 0,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 0,
+        "present_total": 0,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 0,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"state\":[\"Louisiana\"],\"condition_temperature\":[\"colder\"]}",
+      "value": {
+        "state": [
+          "Louisiana"
+        ],
+        "condition_temperature": [
+          "colder"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 23,
+      "seconds": 0.09773133299313486,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-09",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 2,
+        "present_total": 2,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"current_location\":[],\"geographic_poi\":[],\"condition_temperature\":[\"colder\"]}",
+      "value": {
+        "current_location": [],
+        "geographic_poi": [],
+        "condition_temperature": [
+          "colder"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 31,
+      "seconds": 0.12699225009419024,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-10",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"timeRange\":[\"last August\"],\"geographic_poi\":[\"Stunt Ranch Santa Monica Mountains Reserve\"],\"condition_temperature\":[\"colder\"]}",
+      "value": {
+        "timeRange": [
+          "last August"
+        ],
+        "geographic_poi": [
+          "Stunt Ranch Santa Monica Mountains Reserve"
+        ],
+        "condition_temperature": [
+          "colder"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 44,
+      "seconds": 0.16674145800061524,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-11",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 3,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 3,
+        "present_total": 3,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 3,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"state\":[],\"geographic_poi\":[\"Stunt Ranch Santa Monica Mountains Reserve\"],\"current_location\":[]}",
+      "value": {
+        "state": [],
+        "geographic_poi": [
+          "Stunt Ranch Santa Monica Mountains Reserve"
+        ],
+        "current_location": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 34,
+      "seconds": 0.09643245814368129,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-12",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"artist\":[\"Behzad Mirkhani\"],\"music_item\":[\"theme\"],\"sort\":[\"last\"]}",
+      "value": {
+        "artist": [
+          "Behzad Mirkhani"
+        ],
+        "music_item": [
+          "theme"
+        ],
+        "sort": [
+          "last"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 38,
+      "seconds": 0.0911900419741869,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-13",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 3,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 3,
+        "present_total": 3,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 3,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"playlist\":[],\"music_item\":[\"theme\"],\"service\":[]}",
+      "value": {
+        "playlist": [],
+        "music_item": [
+          "theme"
+        ],
+        "service": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 25,
+      "seconds": 0.07062204205431044,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-14",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"artist\":[\"Yui\"],\"music_item\":[\"track\"],\"service\":[\"Vimeo\"]}",
+      "value": {
+        "artist": [
+          "Yui"
+        ],
+        "music_item": [
+          "track"
+        ],
+        "service": [
+          "Vimeo"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 33,
+      "seconds": 0.1308659170754254,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-15",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 3,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 3,
+        "present_total": 3,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 3,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"playlist\":[],\"sort\":[]}",
+      "value": {
+        "playlist": [],
+        "sort": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 14,
+      "seconds": 0.05398241593502462,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-16",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 0,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 0,
+        "present_total": 0,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 0,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"rating_value\":[\"3\"],\"rating_unit\":[\"stars\"],\"object_name\":[\"Things Fall Apart\"]}",
+      "value": {
+        "rating_value": [
+          "3"
+        ],
+        "rating_unit": [
+          "stars"
+        ],
+        "object_name": [
+          "Things Fall Apart"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 35,
+      "seconds": 0.1139006249140948,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-17",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 3,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 3,
+        "present_total": 3,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 3,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"object_name\":[\"Things Fall Apart\"],\"object_select\":[],\"object_type\":[]}",
+      "value": {
+        "object_name": [
+          "Things Fall Apart"
+        ],
+        "object_select": [],
+        "object_type": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 29,
+      "seconds": 0.08503687498159707,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-18",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"rating_value\":[\"1\"],\"object_name\":[]}",
+      "value": {
+        "rating_value": [
+          "1"
+        ],
+        "object_name": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 19,
+      "seconds": 0.060387207893654704,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-19",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 1,
+        "absent_total": 1,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"object_part_of_series_type\":[],\"object_name\":[]}",
+      "value": {
+        "object_part_of_series_type": [],
+        "object_name": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 23,
+      "seconds": 0.09527637506835163,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-20",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 0,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 0,
+        "present_total": 0,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 0,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"object_name\":[\"Happiness Was Free\"]}",
+      "value": {
+        "object_name": [
+          "Happiness Was Free"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 16,
+      "seconds": 0.05649704090319574,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-21",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 1,
+        "fields": 1,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"object_type\":[],\"object_name\":[\"Happiness Was Free\"]}",
+      "value": {
+        "object_type": [],
+        "object_name": [
+          "Happiness Was Free"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 23,
+      "seconds": 0.059290583012625575,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-22",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 1,
+        "absent_total": 1,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"object_name\":[\"Martin Morning\"],\"object_type\":[\"saga\"]}",
+      "value": {
+        "object_name": [
+          "Martin Morning"
+        ],
+        "object_type": [
+          "saga"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 24,
+      "seconds": 0.0722098748665303,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-23",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 2,
+        "present_total": 2,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"object_name\":[\"Martin Morning\"]}",
+      "value": {
+        "object_name": [
+          "Martin Morning"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 13,
+      "seconds": 0.056712874909862876,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-24",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 1,
+        "fields": 1,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"movie_type\":[\"animated movies\"],\"spatial_relation\":[\"in the neighborhood\"]}",
+      "value": {
+        "movie_type": [
+          "animated movies"
+        ],
+        "spatial_relation": [
+          "in the neighborhood"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 29,
+      "seconds": 0.11536691710352898,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-25",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 2,
+        "present_total": 2,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"movie_name\":[],\"timeRange\":[],\"movie_type\":[\"animated movies\"]}",
+      "value": {
+        "movie_name": [],
+        "timeRange": [],
+        "movie_type": [
+          "animated movies"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 32,
+      "seconds": 0.12519749999046326,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-26",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"timeRange\":[\"twenty two o'clock\"],\"movie_name\":[\"Mabel’s Dramatic Career\"]}",
+      "value": {
+        "timeRange": [
+          "twenty two o'clock"
+        ],
+        "movie_name": [
+          "Mabel’s Dramatic Career"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 36,
+      "seconds": 0.1416236248333007,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-27",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 2,
+        "present_total": 2,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"timeRange\":[\"twenty two o'clock\"],\"spatial_relation\":[],\"object_location_type\":[]}",
+      "value": {
+        "timeRange": [
+          "twenty two o'clock"
+        ],
+        "spatial_relation": [],
+        "object_location_type": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 35,
+      "seconds": 0.14057387504726648,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-28",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"track\":[\"Blue Lantern\"],\"artist\":[\"Mira Sol\"],\"album\":[]}",
+      "value": {
+        "track": [
+          "Blue Lantern"
+        ],
+        "artist": [
+          "Mira Sol"
+        ],
+        "album": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 29,
+      "seconds": 0.11759562487713993,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-29",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 2,
+        "present_total": 2,
+        "absent_correct": 1,
+        "absent_total": 1,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"restaurant_name\":[\"River Moon\"],\"city\":[\"Denver\"],\"party_size_number\":[\"four\"],\"timeRange\":[\"tomorrow\"]}",
+      "value": {
+        "restaurant_name": [
+          "River Moon"
+        ],
+        "city": [
+          "Denver"
+        ],
+        "party_size_number": [
+          "four"
+        ],
+        "timeRange": [
+          "tomorrow"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 49,
+      "seconds": 0.17598291696049273,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-30",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 4,
+        "fields": 4,
+        "tp": 4,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 4,
+        "present_total": 4,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 4,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"city\":[\"Ose\"],\"timeRange\":[\"Friday\"],\"country\":[]}",
+      "value": {
+        "city": [
+          "Ose"
+        ],
+        "timeRange": [
+          "Friday"
+        ],
+        "country": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 26,
+      "seconds": 0.0654072081670165,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-31",
+      "checks": {
+        "exact": 0,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 3,
+        "tp": 1,
+        "fp": 1,
+        "fn": 1,
+        "present_correct": 1,
+        "present_total": 2,
+        "absent_correct": 1,
+        "absent_total": 1,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 1,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"object_name\":[\"Glass Harbor\"],\"rating_value\":[\"four\"],\"best_rating\":[],\"rating_unit\":[\"stars\"]}",
+      "value": {
+        "object_name": [
+          "Glass Harbor"
+        ],
+        "rating_value": [
+          "four"
+        ],
+        "best_rating": [],
+        "rating_unit": [
+          "stars"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 42,
+      "seconds": 0.11358241690322757,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-32",
+      "checks": {
+        "exact": 0,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 4,
+        "tp": 3,
+        "fp": 0,
+        "fn": 1,
+        "present_correct": 3,
+        "present_total": 4,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 1,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 3,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"timeRange\":[],\"location_name\":[\"Maple Cinema\"],\"movie_name\":[\"Silver Orchard\"]}",
+      "value": {
+        "timeRange": [],
+        "location_name": [
+          "Maple Cinema"
+        ],
+        "movie_name": [
+          "Silver Orchard"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 37,
+      "seconds": 0.10278037493117154,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-33",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 2,
+        "present_total": 2,
+        "absent_correct": 1,
+        "absent_total": 1,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"entity_name\":[\"Night Steps\"],\"playlist\":[\"Evening Walks\"]}",
+      "value": {
+        "entity_name": [
+          "Night Steps"
+        ],
+        "playlist": [
+          "Evening Walks"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 27,
+      "seconds": 0.08685358287766576,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-34",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 2,
+        "present_total": 2,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"artist\":[\"Mira Sol\",\"Theo Lane\"],\"album\":[]}",
+      "value": {
+        "artist": [
+          "Mira Sol",
+          "Theo Lane"
+        ],
+        "album": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 24,
+      "seconds": 0.10270537505857646,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-35",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 1,
+        "absent_total": 1,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"city\":[\"Bergen\",\"Orange\",\"Lake\"],\"country\":[]}",
+      "value": {
+        "city": [
+          "Bergen",
+          "Orange",
+          "Lake"
+        ],
+        "country": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 24,
+      "seconds": 0.09942620806396008,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-36",
+      "checks": {
+        "exact": 0,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 1,
+        "fields": 2,
+        "tp": 1,
+        "fp": 2,
+        "fn": 1,
+        "present_correct": 0,
+        "present_total": 1,
+        "absent_correct": 1,
+        "absent_total": 1,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 2,
+        "predicted_values": 3,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"cuisine\":[],\"city\":[\"Boston\"],\"restaurant_name\":[]}",
+      "value": {
+        "cuisine": [],
+        "city": [
+          "Boston"
+        ],
+        "restaurant_name": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 26,
+      "seconds": 0.1082186670973897,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-37",
+      "checks": {
+        "exact": 0,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 3,
+        "tp": 1,
+        "fp": 0,
+        "fn": 1,
+        "present_correct": 1,
+        "present_total": 2,
+        "absent_correct": 1,
+        "absent_total": 1,
+        "false_abstention": 1,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"track\":[],\"artist\":[\"Mira Sol\"],\"album\":[\"Quiet Signals\"]}",
+      "value": {
+        "track": [],
+        "artist": [
+          "Mira Sol"
+        ],
+        "album": [
+          "Quiet Signals"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 30,
+      "seconds": 0.12260229187086225,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-38",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 2,
+        "present_total": 2,
+        "absent_correct": 1,
+        "absent_total": 1,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"object_name\":[\"Paper Bridge\"],\"rating_value\":[]}",
+      "value": {
+        "object_name": [
+          "Paper Bridge"
+        ],
+        "rating_value": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 20,
+      "seconds": 0.0860819579102099,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-39",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 1,
+        "absent_total": 1,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"artist\":[],\"timeRange\":[],\"country\":[]}",
+      "value": {
+        "artist": [],
+        "timeRange": [],
+        "country": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 20,
+      "seconds": 0.08448487496934831,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "FIELD-40",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 0,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 0,
+        "present_total": 0,
+        "absent_correct": 3,
+        "absent_total": 3,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 0,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"selected_playlist\":[\"night vibes\"],\"selected_artist\":[],\"selected_playlist_owner\":[]}",
+      "value": {
+        "selected_playlist": [
+          "night vibes"
+        ],
+        "selected_artist": [],
+        "selected_playlist_owner": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 36,
+      "seconds": 0.14466900005936623,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:AddToPlaylist:train:80:view1:native:alias",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"playlist_owner\":null,\"artist\":\"gn\",\"playlist\":\"night vibes\"}",
+      "value": {
+        "playlist_owner": null,
+        "artist": "gn",
+        "playlist": "night vibes"
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 26,
+      "seconds": 0.09343320806510746,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:AddToPlaylist:train:80:view1:native:nullable",
+      "checks": {
+        "exact": 0,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 3,
+        "tp": 1,
+        "fp": 1,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 1,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 1,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"selected_restaurant_name\":[\"mughlai restaurant\"],\"selected_timeRange\":[\"Nov. the 1st\"],\"selected_sort\":[]}",
+      "value": {
+        "selected_restaurant_name": [
+          "mughlai restaurant"
+        ],
+        "selected_timeRange": [
+          "Nov. the 1st"
+        ],
+        "selected_sort": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 47,
+      "seconds": 0.16156229190528393,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:BookRestaurant:train:684:view1:native:alias",
+      "checks": {
+        "exact": 0,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 3,
+        "tp": 1,
+        "fp": 1,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 1,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 1,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"restaurant_name\":\"mughlai restaurant\",\"timeRange\":\"Nov. the 1st\",\"sort\":null}",
+      "value": {
+        "restaurant_name": "mughlai restaurant",
+        "timeRange": "Nov. the 1st",
+        "sort": null
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 34,
+      "seconds": 0.13589041586965322,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:BookRestaurant:train:684:view1:native:nullable",
+      "checks": {
+        "exact": 0,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 3,
+        "tp": 1,
+        "fp": 1,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 1,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 1,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"selected_condition_temperature\":[\"colder\"],\"selected_state\":[\"Louisiana\"]}",
+      "value": {
+        "selected_condition_temperature": [
+          "colder"
+        ],
+        "selected_state": [
+          "Louisiana"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 29,
+      "seconds": 0.10488454182632267,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:GetWeather:train:109:view0:native:alias",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 2,
+        "present_total": 2,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"state\":\"Louisiana\",\"condition_temperature\":\"colder\"}",
+      "value": {
+        "state": "Louisiana",
+        "condition_temperature": "colder"
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 19,
+      "seconds": 0.08123483415693045,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:GetWeather:train:109:view0:native:nullable",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 2,
+        "present_total": 2,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"selected_playlist\":[],\"selected_music_item\":[\"theme\"],\"selected_service\":[]}",
+      "value": {
+        "selected_playlist": [],
+        "selected_music_item": [
+          "theme"
+        ],
+        "selected_service": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 34,
+      "seconds": 0.14011824992485344,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:PlayMusic:train:801:view1:native:alias",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"playlist\":null,\"music_item\":\"theme\",\"service\":null}",
+      "value": {
+        "playlist": null,
+        "music_item": "theme",
+        "service": null
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 23,
+      "seconds": 0.09768870798870921,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:PlayMusic:train:801:view1:native:nullable",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"selected_rating_value\":[\"2\"],\"selected_object_select\":[\"current\"]}",
+      "value": {
+        "selected_rating_value": [
+          "2"
+        ],
+        "selected_object_select": [
+          "current"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 29,
+      "seconds": 0.11820799997076392,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:RateBook:train:293:view0:native:alias",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 2,
+        "present_total": 2,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"object_select\":\"current\",\"rating_value\":\"2\"}",
+      "value": {
+        "object_select": "current",
+        "rating_value": "2"
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 19,
+      "seconds": 0.0810724999755621,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:RateBook:train:293:view0:native:nullable",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 2,
+        "tp": 2,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 2,
+        "present_total": 2,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 2,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"selected_object_name\":[\"Playlist: The Very Best of Dolly Parton\"]}",
+      "value": {
+        "selected_object_name": [
+          "Playlist: The Very Best of Dolly Parton"
+        ]
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 25,
+      "seconds": 0.10242487490177155,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:SearchCreativeWork:train:602:view1:native:alias",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 1,
+        "fields": 1,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"object_name\":\"Playlist: The Very Best of Dolly Parton\"}",
+      "value": {
+        "object_name": "Playlist: The Very Best of Dolly Parton"
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 20,
+      "seconds": 0.0858239580411464,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:SearchCreativeWork:train:602:view1:native:nullable",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 1,
+        "fields": 1,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 0,
+        "absent_total": 0,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"selected_movie_type\":[\"animated movies\"],\"selected_movie_name\":[],\"selected_location_name\":[]}",
+      "value": {
+        "selected_movie_type": [
+          "animated movies"
+        ],
+        "selected_movie_name": [],
+        "selected_location_name": []
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 42,
+      "seconds": 0.17146354098804295,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:SearchScreeningEvent:train:897:view1:native:alias",
+      "checks": {
+        "exact": 1,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 3,
+        "fields": 3,
+        "tp": 1,
+        "fp": 0,
+        "fn": 0,
+        "present_correct": 1,
+        "present_total": 1,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     },
     {
+      "text": "{\"movie_name\":null,\"movie_type\":\"movies\",\"location_name\":null}",
+      "value": {
+        "movie_name": null,
+        "movie_type": "movies",
+        "location_name": null
+      },
+      "schema_valid": true,
+      "success": true,
       "stop_reason": "turn_end",
+      "output_tokens": 31,
+      "seconds": 0.1260323328897357,
+      "output_mode": "native-json",
+      "model_calls": 1,
+      "json_assembly": false,
+      "output_repair": false,
+      "constrained": false,
+      "protocol": {
+        "temperature": 0,
+        "repetition_penalty": 1,
+        "max_tokens": 256,
+        "serialization": "native-sft-chat-v1"
+      },
+      "id": "native:snips:SearchScreeningEvent:train:897:view1:native:nullable",
+      "checks": {
+        "exact": 0,
+        "json_parse_ok": 1,
+        "schema_ok": 1,
+        "field_correct": 2,
+        "fields": 3,
+        "tp": 0,
+        "fp": 1,
+        "fn": 1,
+        "present_correct": 0,
+        "present_total": 1,
+        "absent_correct": 2,
+        "absent_total": 2,
+        "false_abstention": 0,
+        "missed_abstention": 0,
+        "unsupported_values": 0,
+        "predicted_values": 1,
+        "natural_stop": 1,
+        "token_limited": 0,
+        "nonrepetitive_nonempty": 1
+      },
+      "saved_text_and_stop_match": true
     }
   ],
+  "model_calls_per_request": 1,
+  "json_assembly": false,
+  "output_repair": false,
+  "constrained_decoding": false,
+  "scope": "Stock MLX standalone packaging parity on the observed 54 playground fixtures; not a fresh blind accuracy test or full 1,444-request cross-runtime parity.",
+  "elapsed_seconds": 5.9345723749138415
 }

JSONLITE_RESULTS.json CHANGED Viewed

The diff for this file is too large to render. See raw diff

PACKAGING_VERIFICATION.json CHANGED Viewed

@@ -1,491 +1,150 @@
 {
   "model": "OpenSML-JSONLite-175M-Instruct",
   "date": "2026-10-14",
-  "selected_stage": "sft-100",
-  "step": 469,
   "weight_file": "model.safetensors",
   "weight_bytes": 351242342,
-  "weight_sha256": "656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4",
-  "native_bundle_sha256": "4ea6280d2e60da34e59757a4d6ab95aaf7811050c68c047cf5b98d2f001e9b78",
-  "runtime": "stock MLX0.32.3, Apple Silicon, Python3.11",
-  "native_generation_verified": {
-    "constrained_decoding": false,
-    "correct_against_reference": 42,
-    "json_assembly": false,
-    "model": "OpenSML-JSONLite-175M-Instruct",
-    "model_calls_per_request": 1,
-    "output_mode": "native-json",
-    "output_repair": false,
-    "outputs": [
-      {
-        "correct": false,
-        "id": "FIELD-01",
-        "index": 1,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"playlist_owner\":[],\"playlist\":[\"Rock Gaming\"],\"entity_name\":[\"Mariem Hassan\"]}"
-      },
-      {
-        "correct": false,
-        "id": "FIELD-02",
-        "index": 2,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"music_item\":[],\"artist\":[\"Mariem Hassan\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-03",
-        "index": 3,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"playlist_owner\":[\"my\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-04",
-        "index": 4,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"artist\":[],\"entity_name\":[],\"playlist\":[\"Digster Future Hits\"]}"
-      },
-      {
-        "correct": false,
-        "id": "FIELD-05",
-        "index": 5,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"state\":[\"AR\"],\"party_size_number\":[\"nine\"],\"state\":[\"AR\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-06",
-        "index": 6,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"party_size_number\":[\"nine\"],\"timeRange\":[],\"sort\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-07",
-        "index": 7,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"sort\":[\"best\"],\"city\":[\"Wrigley\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-08",
-        "index": 8,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"timeRange\":[],\"poi\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-09",
-        "index": 9,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"condition_temperature\":[\"colder\"],\"state\":[\"Louisiana\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-10",
-        "index": 10,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"current_location\":[],\"geographic_poi\":[],\"condition_temperature\":[\"colder\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-11",
-        "index": 11,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"timeRange\":[\"last August\"],\"geographic_poi\":[\"Stunt Ranch Santa Monica Mountains Reserve\"],\"condition_temperature\":[\"colder\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-12",
-        "index": 12,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"current_location\":[],\"geographic_poi\":[\"Stunt Ranch Santa Monica Mountains Reserve\"],\"state\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-13",
-        "index": 13,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"artist\":[\"Behzad Mirkhani\"],\"music_item\":[\"theme\"],\"sort\":[\"last\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-14",
-        "index": 14,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"playlist\":[],\"music_item\":[\"theme\"],\"service\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-15",
-        "index": 15,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"artist\":[\"Yui\"],\"music_item\":[\"track\"],\"service\":[\"Vimeo\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-16",
-        "index": 16,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"sort\":[],\"playlist\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-17",
-        "index": 17,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"rating_value\":[\"3\"],\"rating_unit\":[\"stars\"],\"object_name\":[\"Things Fall Apart\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-18",
-        "index": 18,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"object_name\":[\"Things Fall Apart\"],\"object_select\":[],\"object_type\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-19",
-        "index": 19,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"rating_value\":[\"1\"],\"object_name\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-20",
-        "index": 20,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"object_part_of_series_type\":[],\"object_name\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-21",
-        "index": 21,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"object_name\":[\"Happiness Was Free\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-22",
-        "index": 22,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"object_type\":[],\"object_name\":[\"Happiness Was Free\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-23",
-        "index": 23,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"object_name\":[\"Martin Morning\"],\"object_type\":[\"saga\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-24",
-        "index": 24,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"object_name\":[\"Martin Morning\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-25",
-        "index": 25,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"movie_type\":[\"animated movies\"],\"spatial_relation\":[\"in the neighborhood\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-26",
-        "index": 26,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"movie_name\":[],\"timeRange\":[],\"movie_type\":[\"animated movies\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-27",
-        "index": 27,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"timeRange\":[\"twenty two o'clock\"],\"movie_name\":[\"Mabel\u2019s Dramatic Career\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-28",
-        "index": 28,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"timeRange\":[\"twenty two o'clock\"],\"spatial_relation\":[],\"object_location_type\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-29",
-        "index": 29,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"album\":[],\"artist\":[\"Mira Sol\"],\"track\":[\"Blue Lantern\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-30",
-        "index": 30,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"restaurant_name\":[\"River Moon\"],\"city\":[\"Denver\"],\"party_size_number\":[\"four\"],\"timeRange\":[\"tomorrow\"]}"
-      },
-      {
-        "correct": false,
-        "id": "FIELD-31",
-        "index": 31,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"country\":[],\"timeRange\":[\"Friday\"],\"city\":[\"Odense\"]}"
-      },
-      {
-        "correct": false,
-        "id": "FIELD-32",
-        "index": 32,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"object_name\":[\"Glass Harbor\"],\"rating_value\":[\"four\"],\"best_rating\":[],\"rating_unit\":[\"stars\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-33",
-        "index": 33,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"timeRange\":[],\"location_name\":[\"Maple Cinema\"],\"movie_name\":[\"Silver Orchard\"]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-34",
-        "index": 34,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"entity_name\":[\"Night Steps\"],\"playlist\":[\"Evening Walks\"]}"
-      },
-      {
-        "correct": false,
-        "id": "FIELD-35",
-        "index": 35,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"artist\":[\"Mira Sol and Theo Lane\"],\"album\":[]}"
-      },
-      {
-        "correct": false,
-        "id": "FIELD-36",
-        "index": 36,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"city\":[\"Bergen\"],\"country\":[]}"
-      },
-      {
-        "correct": false,
-        "id": "FIELD-37",
-        "index": 37,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"cuisine\":[],\"city\":[\"Boston\"],\"restaurant_name\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-38",
-        "index": 38,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"album\":[\"Quiet Signals\"],\"artist\":[\"Mira Sol\"],\"track\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-39",
-        "index": 39,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"object_name\":[\"Paper Bridge\"],\"rating_value\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "FIELD-40",
-        "index": 40,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"country\":[],\"timeRange\":[],\"artist\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "native:snips:AddToPlaylist:train:80:view1:native:alias",
-        "index": 41,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"selected_playlist\":[\"night vibes\"],\"selected_artist\":[],\"selected_playlist_owner\":[]}"
-      },
-      {
-        "correct": false,
-        "id": "native:snips:AddToPlaylist:train:80:view1:native:nullable",
-        "index": 42,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"playlist_owner\":null,\"artist\":\"gn\",\"playlist\":night vibes\"}"
-      },
-      {
-        "correct": false,
-        "id": "native:snips:BookRestaurant:train:684:view1:native:alias",
-        "index": 43,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"selected_restaurant_name\":[\"mughlai restaurant\"],\"selected_timeRange\":[\"Nov. the 1st\"],\"selected_sort\":[]}"
-      },
-      {
-        "correct": false,
-        "id": "native:snips:BookRestaurant:train:684:view1:native:nullable",
-        "index": 44,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"restaurant_name\":\"Mughlai Restaurant\",\"timeRange\":\"Nov. the 1st\",\"sort\":null}"
-      },
-      {
-        "correct": true,
-        "id": "native:snips:GetWeather:train:109:view0:native:alias",
-        "index": 45,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"selected_condition_temperature\":[\"colder\"],\"selected_state\":[\"Louisiana\"]}"
-      },
-      {
-        "correct": true,
-        "id": "native:snips:GetWeather:train:109:view0:native:nullable",
-        "index": 46,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"state\":\"Louisiana\",\"condition_temperature\":\"colder\"}"
-      },
-      {
-        "correct": true,
-        "id": "native:snips:PlayMusic:train:801:view1:native:alias",
-        "index": 47,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"selected_playlist\":[],\"selected_music_item\":[\"theme\"],\"selected_service\":[]}"
-      },
-      {
-        "correct": true,
-        "id": "native:snips:PlayMusic:train:801:view1:native:nullable",
-        "index": 48,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"playlist\":null,\"music_item\":\"theme\",\"service\":null}"
-      },
-      {
-        "correct": true,
-        "id": "native:snips:RateBook:train:293:view0:native:alias",
-        "index": 49,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"selected_rating_value\":[\"2\"],\"selected_object_select\":[\"current\"]}"
-      },
-      {
-        "correct": true,
-        "id": "native:snips:RateBook:train:293:view0:native:nullable",
-        "index": 50,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"object_select\":\"current\",\"rating_value\":\"2\"}"
-      },
-      {
-        "correct": true,
-        "id": "native:snips:SearchCreativeWork:train:602:view1:native:alias",
-        "index": 51,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"selected_object_name\":[\"Playlist: The Very Best of Dolly Parton\"]}"
-      },
-      {
-        "correct": true,
-        "id": "native:snips:SearchCreativeWork:train:602:view1:native:nullable",
-        "index": 52,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"object_name\":\"Playlist: The Very Best of Dolly Parton\"}"
-      },
-      {
-        "correct": true,
-        "id": "native:snips:SearchScreeningEvent:train:897:view1:native:alias",
-        "index": 53,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"selected_movie_type\":[\"animated movies\"],\"selected_movie_name\":[],\"selected_location_name\":[]}"
-      },
-      {
-        "correct": false,
-        "id": "native:snips:SearchScreeningEvent:train:897:view1:native:nullable",
-        "index": 54,
-        "raw_parity": true,
-        "stop_reason": "turn_end",
-        "text": "{\"movie_name\":null,\"movie_type\":\"movies\",\"location_name\":null}"
       }
     ],
-    "raw_text_and_stop_matches": 54,
-    "requests": 54,
-    "scope": "Clean stock-MLX packaging parity on observed playground fixtures; not full-heldout runtime parity or fresh blind accuracy.",
-    "step": 469,
-    "versions": {
-      "jsonschema": "4.26.0",
-      "mlx": "0.32.3",
-      "numpy": "2.4.2",
-      "tokenizers": "0.22.2"
-    },
-    "weight_sha256": "656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4"
-  },
-  "root_runner_verification": {
-    "command": "python inference.py --example 29 --json",
-    "default_weights": true,
-    "source_bundle_integrity": true,
-    "frozen_tokenizer_integrity": true,
-    "exact": true,
-    "raw_text": "{\"album\":[],\"artist\":[\"Mira Sol\"],\"track\":[\"Blue Lantern\"]}",
-    "value": {
-      "album": [],
-      "artist": [
-        "Mira Sol"
-      ],
-      "track": [
-        "Blue Lantern"
-      ]
-    },
-    "model_calls": 1
   },
-  "tokenizer_root_matches_bundle": true,
-  "scope": "54-case raw-output clean-runtime parity; not fresh blind or full-heldout cross-runtime validation.",
-  "model_card_examples": [
-    29,
-    30,
-    40,
-    48
-  ]
 }

 {
   "model": "OpenSML-JSONLite-175M-Instruct",
   "date": "2026-10-14",
+  "checkpoint": "main",
   "weight_file": "model.safetensors",
   "weight_bytes": 351242342,
+  "weight_sha256": "655fb5cfba009163cd8a7f7fe7b170b998bb452c217b1b473cfe10b33de0ccd1",
+  "native_bundle_sha256": "a1b63ed6286d6c4f8bc4f7526f4a11fc46f936830daaac22224bfa0c4892755b",
+  "runtime": "stock MLX 0.32.3; Apple Silicon; Python 3.11",
+  "requests": 54,
+  "raw_runtime_matches": 54,
+  "complete_correct": 44,
+  "schema_valid": 54,
+  "root_runner_verification": {
+    "passed": true,
+    "examples": [
+      {
+        "example": 29,
+        "result": {
+          "text": "{\"track\":[\"Blue Lantern\"],\"artist\":[\"Mira Sol\"],\"album\":[]}",
+          "value": {
+            "track": [
+              "Blue Lantern"
+            ],
+            "artist": [
+              "Mira Sol"
+            ],
+            "album": []
+          },
+          "schema_valid": true,
+          "success": true,
+          "stop_reason": "turn_end",
+          "output_tokens": 29,
+          "seconds": 0.09680337482132018,
+          "output_mode": "native-json",
+          "model_calls": 1,
+          "json_assembly": false,
+          "output_repair": false,
+          "constrained": false,
+          "protocol": {
+            "temperature": 0,
+            "repetition_penalty": 1,
+            "max_tokens": 256,
+            "serialization": "native-sft-chat-v1"
+          },
+          "exact": true
+        }
+      },
+      {
+        "example": 30,
+        "result": {
+          "text": "{\"restaurant_name\":[\"River Moon\"],\"city\":[\"Denver\"],\"party_size_number\":[\"four\"],\"timeRange\":[\"tomorrow\"]}",
+          "value": {
+            "restaurant_name": [
+              "River Moon"
+            ],
+            "city": [
+              "Denver"
+            ],
+            "party_size_number": [
+              "four"
+            ],
+            "timeRange": [
+              "tomorrow"
+            ]
+          },
+          "schema_valid": true,
+          "success": true,
+          "stop_reason": "turn_end",
+          "output_tokens": 49,
+          "seconds": 0.14814749988727272,
+          "output_mode": "native-json",
+          "model_calls": 1,
+          "json_assembly": false,
+          "output_repair": false,
+          "constrained": false,
+          "protocol": {
+            "temperature": 0,
+            "repetition_penalty": 1,
+            "max_tokens": 256,
+            "serialization": "native-sft-chat-v1"
+          },
+          "exact": true
+        }
+      },
+      {
+        "example": 35,
+        "result": {
+          "text": "{\"artist\":[\"Mira Sol\",\"Theo Lane\"],\"album\":[]}",
+          "value": {
+            "artist": [
+              "Mira Sol",
+              "Theo Lane"
+            ],
+            "album": []
+          },
+          "schema_valid": true,
+          "success": true,
+          "stop_reason": "turn_end",
+          "output_tokens": 24,
+          "seconds": 0.09080504113808274,
+          "output_mode": "native-json",
+          "model_calls": 1,
+          "json_assembly": false,
+          "output_repair": false,
+          "constrained": false,
+          "protocol": {
+            "temperature": 0,
+            "repetition_penalty": 1,
+            "max_tokens": 256,
+            "serialization": "native-sft-chat-v1"
+          },
+          "exact": true
+        }
+      },
+      {
+        "example": 48,
+        "result": {
+          "text": "{\"playlist\":null,\"music_item\":\"theme\",\"service\":null}",
+          "value": {
+            "playlist": null,
+            "music_item": "theme",
+            "service": null
+          },
+          "schema_valid": true,
+          "success": true,
+          "stop_reason": "turn_end",
+          "output_tokens": 23,
+          "seconds": 0.0877893331926316,
+          "output_mode": "native-json",
+          "model_calls": 1,
+          "json_assembly": false,
+          "output_repair": false,
+          "constrained": false,
+          "protocol": {
+            "temperature": 0,
+            "repetition_penalty": 1,
+            "max_tokens": 256,
+            "serialization": "native-sft-chat-v1"
+          },
+          "exact": true
+        }
       }
     ],
+    "scope": "Four documented examples via packaged archive bootstrap, no assembly or repair."
   },
+  "json_assembly": false,
+  "output_repair": false,
+  "tokenizer_root_matches_bundle": true
 }

README.md CHANGED Viewed

@@ -27,73 +27,61 @@ datasets:
 **Author:** William Zebrowski · **175M parameters**
-[Run locally](RUN_GUIDE.md) · [Technical report](TECHNICAL_REPORT.md) · [Results](JSONLITE_RESULTS.json) · [Provenance](checkpoint_provenance.json) · [Inference verification](INFERENCE_VERIFICATION.json)
-**A small language model that generates JSON directly from source text and a supplied JSON Schema.** It produces the entire object in one native autoregressive request. There is no fieldwise JSON assembler, output repair or constrained decoder in this release. Schema validation happens after generation and cannot establish that extracted values are correct.
-This update replaces the earlier extraction-plus-assembly weights and runner with a native source/schema → JSON fine-tune. The pretrained ancestor is [OpenSML-Base-175M](https://huggingface.co/OpenSML/OpenSML-Base-175M); the model adds evidence QA, exact field extraction, focused extraction and a final directly supervised JSON stage. Full lineage and exact artifact hashes are in the report.
 ## Run locally
-The repository contains BF16 `model.safetensors`, tokenizer/configuration, native MLX source archive, runnable entry point, examples and verification/results records. This is a custom architecture; use the supplied runner rather than Transformers AutoModel or mlx-lm auto-loading. Verified platform: Apple Silicon macOS, Python 3.11 and stock MLX 0.32.3.
 ```bash
 python3 -m venv .venv
 source .venv/bin/activate
 python -m pip install huggingface_hub
-# Authenticate if access requires it: hf auth login
 hf download OpenSML/OpenSML-JSONLite-175M --local-dir OpenSML-JSONLite-175M
 cd OpenSML-JSONLite-175M
 python -m pip install -r requirements.txt
-python inference.py --example 29
 python inference.py --text-file request.txt --schema-file schema.json
-# Optional: raw text plus validation/stop metadata and example correctness
-python inference.py --example 29 --json
 ```
-Default stdout is the model's unmodified generated text. The fixed extraction instruction and native prompt template are supplied by the runner. Example answers are checked afterward and never passed to generation. After downloading and installing dependencies, inference runs offline. The root runner verifies the source archive and selected weight SHA-256. Tokenizer companions are in the bundle; optimizer state is not needed for inference.
-## Model and inputs
-| Setting | Value |
-| --- | --- |
-| Parameters | 175,611,648 |
-| Layers / hidden / FFN width | 24 / 768 / 2,048 |
-| Attention | GQA, 12 query heads / 4 KV heads |
-| Normalization / MLP / positions | RMSNorm + Q/K norm / SwiGLU / RoPE 10,000 |
-| Vocabulary / context | Frozen 32,000-token byte-BPE / 2,048 shared tokens |
-| Embeddings / biases | Tied input-output / no linear biases |
-| Weights | BF16, unquantized safetensors |
-| Output | One native JSON object per request |
-| Default decoding | Temperature 0, repetition penalty 1, max 256 output tokens |
-Supply source text containing the facts and a closed flat object schema with 1–8 required properties and `additionalProperties: false`. Supported property types are strings, nullable strings (`["string","null"]`) and arrays of strings. Clear field descriptions help specify semantics. Missing arrays target `[]`; missing nullable strings target `null`. Strings preserve source case/spelling and arrays preserve source order; number/date normalization is not trained. Overlong inputs reject without truncation.
-The run is trained primarily on seven SNIPS domains and 39 field definitions. New names/descriptions are passed to the model directly, but reliable unfamiliar-domain extraction has not been established. Nested schemas, numeric/boolean outputs, enums, references and advanced constraints are outside the verified runner contract.
-## Verified interaction examples
-These four examples passed the playground reference check with the selected native JSON weights. Their raw outputs also matched the standalone MLX runner. They demonstrate observed successes, not a guarantee on other inputs. Decode greedily: temperature 0, repetition penalty 1, at most 256 output tokens.
-### Instruction and input format
-Use this exact instruction for every example (the provided runner inserts it automatically):
 ```text
 Extract information from the supplied source_text according to the supplied JSON Schema. Return exactly one JSON object, with no explanation or Markdown. Copy values exactly, preserving case and spelling; do not normalize dates or numbers. Include every required key and no other keys. For string arrays, include all matching source spans in source order, or [] when absent. For nullable strings, use the exact source span or null when absent.
 ```
-The user message is that instruction, a blank line, then the compact JSON object `{"source_text": <query>, "schema": <schema>}`. Both inputs are provided to one model generation; no expected answer is included. The runner applies the native SFT role tokens and validates the result afterward.
-### Example 1: Track and artist
-**Query / source text**
 ```text
 Play the track Blue Lantern by Mira Sol.
 ```
-**JSON Schema**
 ```json
 {
@@ -130,23 +118,23 @@ Play the track Blue Lantern by Mira Sol.
 }
 ```
-**Actual model output**
 ```json
-{"album":[],"artist":["Mira Sol"],"track":["Blue Lantern"]}
 ```
-Run the packaged example: `python inference.py --example 29`.
-### Example 2: Restaurant booking
-**Query / source text**
 ```text
 Book River Moon in Denver for four people tomorrow.
 ```
-**JSON Schema**
 ```json
 {
@@ -191,76 +179,68 @@ Book River Moon in Denver for four people tomorrow.
 }
 ```
-**Actual model output**
 ```json
 {"restaurant_name":["River Moon"],"city":["Denver"],"party_size_number":["four"],"timeRange":["tomorrow"]}
 ```
-Run the packaged example: `python inference.py --example 30`.
-### Example 3: Missing fields as empty arrays
-**Query / source text**
 ```text
-Please play some music.
 ```
-**JSON Schema**
 ```json
 {
   "type": "object",
   "properties": {
-    "country": {
-      "type": "array",
-      "items": {
-        "type": "string"
-      },
-      "description": "country"
-    },
-    "timeRange": {
       "type": "array",
       "items": {
         "type": "string"
       },
-      "description": "date, time, or time-range expression"
     },
-    "artist": {
       "type": "array",
       "items": {
         "type": "string"
       },
-      "description": "performer or musical artist"
     }
   },
   "required": [
-    "country",
-    "timeRange",
-    "artist"
   ],
   "additionalProperties": false
 }
 ```
-**Actual model output**
 ```json
-{"country":[],"timeRange":[],"artist":[]}
 ```
-Run the packaged example: `python inference.py --example 40`.
-### Example 4: Nullable strings and requested scope
-**Query / source text**
 ```text
 Play the last theme by Behzad Mirkhani.
 ```
-**JSON Schema**
 ```json
 {
@@ -298,79 +278,60 @@ Play the last theme by Behzad Mirkhani.
 }
 ```
-**Actual model output**
 ```json
 {"playlist":null,"music_item":"theme","service":null}
 ```
-Run the packaged example: `python inference.py --example 48`.
-The performer is present in the source but is not requested by this schema. The output includes only the requested music category and missing service/playlist values.
 ## Training
-The 175.6M-parameter base was pretrained from random initialization on **17,300,111,360 lifetime token positions**, using a five-Mac Apple Silicon cluster with Thunderbolt/RDMA.
-| Pretraining source | Intended sampling share |
-| --- | ---: |
-| FineWeb-Edu | 50% |
-| DCLM-Edu | 18% |
-| English FineWiki | 13% |
-| Cosmopedia v2 | 9% |
-| Educational Python | 10% |
-These are configured token-sampling weights, not unique-document counts. Selected-base validation loss was 2.5563359552. Pretraining architecture, tokenizer, source revisions, filtering, scheduler and precision are documented in the report.
-### Training run
 ![Training loss and held-out validation](pretraining-training-loss.png)
-*Training loss and held-out validation during pretraining. Left: raw training loss and exponential moving average. Right: held-out validation loss.*
-### Supervised adaptation
-| Stage | What it teaches | Inherited training |
-| --- | --- | --- |
-| Grounded QA | Answers from evidence, absence, contrasting questions, multipart/corrections | 15,008 presentations through selected update 469 |
-| Requested-field extraction | Exact SNIPS annotated spans, field definitions, missing values | 12,115 requests / 379 updates |
-| Focused extraction | Single-field supervision plus labeled replay | 5,120 presentations / 160 updates |
-| **Native JSON SFT** | **Source text + full schema → JSON assistant targets** | **15,000 presentations / 469 updates** |
-The final JSON stage has **12,115 distinct source requests**: 12,115 canonical-schema examples, 1,443 renamed-key views and 1,442 nullable-string views. Additional views reuse TRAIN-only sources. Original SNIPS chunk annotations are reconstructed, checked against source offsets and schema-validated; upstream annotation fidelity is not a guarantee of perfect semantic golds.
-The final stage runs one full-weight pass on Mac1 with batch 32, microbatch 2, fresh AdamW, peak LR 5e-6/final 1e-6, 20-update warmup then cosine decay, betas (0.9,0.95), epsilon 1e-8, weight decay 0.01 and clipping 1. Only assistant JSON and stop tokens contribute to the loss. It processes 4,634,126 serialized positions; there is no RL or adapter-only training in this chain.
-DEV selection happens at updates 0/100/200/300/469 on 252 fixed requests. The final artifact is selected before the 1,444-request test. TRAIN/DEV/TEST source groups are separate, but these held-out panels have prior research exposure and are not a new blind benchmark.
 ## Native JSON evaluation
-| Metric | Adapted SNIPS held-out, 1,444 requests |
-| --- | ---: |
-| Entire request correct | **1,191/1,444 — 82.48%** |
-| Field accuracy | **91.31%** |
-| Extracted-value precision / recall / F1 | **90.55% / 90.41% / 90.48%** |
-| Missing-field accuracy | **92.46%** |
-| Schema-valid JSON | **99.79%** |
-| False abstentions / unsupported values | 111 / 30 |
-| Token-limited outputs | 0 |
-Canonical string-array schemas alone score **1,156/1,388 (83.29%)**, with 100% schema validity. Extra renamed-key and nullable-string views score **19/28** and **16/28**. The full 54-case local playground scores **42/54 (77.78%)**, including **7/12** previously authored natural examples. These are different panels; do not pool or present them as an external leaderboard.
-All outputs are raw, greedy native model generations: one whole-object request, no constrained decoder, assembly, repair, prompt truncation or reference answer in inputs. Complete correctness requires natural completion, strict JSON/schema validity and exact values. Structural validity is not factual correctness. [Results](JSONLITE_RESULTS.json) and the report contain counts, per-domain performance, protocols and failures.
-![Native JSON correctness on separate research panels](complete-request-exact.svg)
-## Limitations
-The model remains a focused extraction research candidate. Failures include inventing a city, omitting second values, merging two names, missing rating/cuisine information, confusing entity types, incomplete spans, duplicate keys and malformed nullable strings. Broader natural language, corrections, associations, unfamiliar definitions, long documents and complex schemas need further validation. General dialogue, coding, multilingual quality, confidence calibration and prompt-injection robustness are not established. `success` means natural completion and structure validity, not verified factuality or release readiness.
-## Identity and packaging
-Selected weight SHA-256:
-`656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4`
-Weights have 351,242,342 bytes. The frozen tokenizer is unchanged. Clean stock-MLX packaging verification checks raw output text and stop reasons against all 54 actual playground predictions; its 42/54 correctness is the same observed result, not an independent blind test or full-heldout runtime-parity claim. CUDA, CPU-only, quantized and converted exports remain unverified.
-Source code uses Apache-2.0; upstream data/dependencies retain their terms. The source license does not automatically assign a model-weight license. See the included notices and provenance.

 **Author:** William Zebrowski · **175M parameters**
+[Run locally](RUN_GUIDE.md) · [Technical report](TECHNICAL_REPORT.md) · [Results](JSONLITE_RESULTS.json) · [Provenance](checkpoint_provenance.json)
+**Main checkpoint: a small language model that generates complete JSON directly from source text and a supplied JSON Schema.** It makes one autoregressive call; parsing/schema validation follows generation. No JSON assembly, output repair, constrained decoder or secondary model is used.
+This release retains the grounded-QA, exact/focused extraction and native JSON foundation, then adds a focused extraction continuation with original TRAIN replay, matched zero/one/two/three-value examples and frozen-parent distillation. The display name and standard `model.safetensors` filename stay unchanged. The pretrained ancestor is [OpenSML-Base-175M](https://huggingface.co/OpenSML/OpenSML-Base-175M).
 ## Run locally
 ```bash
 python3 -m venv .venv
 source .venv/bin/activate
 python -m pip install huggingface_hub
 hf download OpenSML/OpenSML-JSONLite-175M --local-dir OpenSML-JSONLite-175M
 cd OpenSML-JSONLite-175M
 python -m pip install -r requirements.txt
+python inference.py --example 35
 python inference.py --text-file request.txt --schema-file schema.json
+python inference.py --example 35 --json
 ```
+The runner inserts the fixed instruction/native role serialization. Default stdout is the unmodified generated text; `--json` adds validation/stop metadata. Example golds are checked afterward and never passed to generation. Inference is offline after download/dependency installation. This custom architecture uses the supplied native MLX runner; Transformers AutoModel and mlx-lm auto-loading are not supported exports. Verified platform: Apple Silicon macOS, Python 3.11, stock MLX 0.32.3.
+## Inputs and supported schema
+Supply source text and a flat object JSON Schema with 1–8 required properties and `additionalProperties:false`. Supported properties are strings, nullable strings and arrays of strings. Use clear field descriptions. Copy exact spelling/case and complete spans; preserve source order in arrays. Missing arrays target []; missing nullable strings target null. There is no date/number normalization. Nested objects, numeric/boolean outputs, enums/references/advanced constraints are outside the verified runner contract. Overlong prompts reject without truncation.
+|Model setting|Value|
+|---|---|
+|Parameters|175,611,648|
+|Layers / hidden / FFN|24 /768 /2,048|
+|Attention|GQA: 12  query / 4 KV heads|
+|Normalization / MLP / positions|RMSNorm+Q/K norm /SwiGLU /RoPE10,000|
+|Vocabulary / shared context|Frozen 32,000-byte-BPE /2,048 tokens|
+|Weights|BF16 unquantized safetensors; tied embeddings|
+|Default generation|Temperature 0, repetition penalty 1, max 256 tokens|
+## Verified source/schema/output examples
+These four observed examples are checked against the selected model and standalone runner. They demonstrate successes within a limited research scope; other requests can fail. The runner supplies this exact instruction:
 ```text
 Extract information from the supplied source_text according to the supplied JSON Schema. Return exactly one JSON object, with no explanation or Markdown. Copy values exactly, preserving case and spelling; do not normalize dates or numbers. Include every required key and no other keys. For string arrays, include all matching source spans in source order, or [] when absent. For nullable strings, use the exact source span or null when absent.
 ```
+The user message is the instruction, a blank line, then compact `{"source_text":<query>,"schema":<schema>}`. No expected answer is included.
+### Track and artist
+Source text:
 ```text
 Play the track Blue Lantern by Mira Sol.
 ```
+JSON Schema:
 ```json
 {
 }
 ```
+Actual model output:
 ```json
+{"track":["Blue Lantern"],"artist":["Mira Sol"],"album":[]}
 ```
+Run: `python inference.py --example 29`.
+### Restaurant request
+Source text:
 ```text
 Book River Moon in Denver for four people tomorrow.
 ```
+JSON Schema:
 ```json
 {
 }
 ```
+Actual model output:
 ```json
 {"restaurant_name":["River Moon"],"city":["Denver"],"party_size_number":["four"],"timeRange":["tomorrow"]}
 ```
+Run: `python inference.py --example 30`.
+### Multiple artists
+Source text:
 ```text
+Play songs by Mira Sol and Theo Lane.
 ```
+JSON Schema:
 ```json
 {
   "type": "object",
   "properties": {
+    "artist": {
       "type": "array",
       "items": {
         "type": "string"
       },
+      "description": "performer or musical artist"
     },
+    "album": {
       "type": "array",
       "items": {
         "type": "string"
       },
+      "description": "album name"
     }
   },
   "required": [
+    "artist",
+    "album"
   ],
   "additionalProperties": false
 }
 ```
+Actual model output:
 ```json
+{"artist":["Mira Sol","Theo Lane"],"album":[]}
 ```
+Run: `python inference.py --example 35`.
+### Nullable fields
+Source text:
 ```text
 Play the last theme by Behzad Mirkhani.
 ```
+JSON Schema:
 ```json
 {
 }
 ```
+Actual model output:
 ```json
 {"playlist":null,"music_item":"theme","service":null}
 ```
+Run: `python inference.py --example 48`.
 ## Training
+The base was pretrained from scratch on 17,300,111,360 token positions using five Apple Silicon Macs with Thunderbolt/RDMA distributed training. Intended mixture: 50% FineWeb-Edu, 18% DCLM-Edu, 13% English FineWiki, 9% Cosmopedia v2,10% educational Python. These are configured sampling shares, not unique-document counts. Base held-out validation loss: 2.5563359552.
 ![Training loss and held-out validation](pretraining-training-loss.png)
+*Training loss and held-out validation during pretraining.*
+|Adaptation stage|What is inherited|
+|---|---|
+|Grounded evidence QA|15,008 presentations through the selected ancestor|
+|Requested-field extraction|12,115 SNIPS requests|
+|Focused extraction plus replay|5,120 presentations|
+|Native JSON foundation|15,000 source/schema/JSON presentations|
+|Additional extraction SFT|2,048 inherited presentations: 1,536 TRAIN replay + 512 controlled cardinality views|
+The additional stage completed a 3,072-presentation experiment; the main checkpoint inherits only its selected 2,048-presentation prefix and 600,558 serialized positions. All stages update full model weights. The add-on uses batch 32/micro 2, fresh AdamW, 6-update warmup, peak 2e-6 and a 96-update cosine schedule toward 2e-7. On replay assistant tokens it adds KL from the frozen native JSON foundation; new rows use gold JSON CE alone. No teacher, RL, LoRA or adapter is needed at inference. Exact settings, data separation and stage hashes are in the report/provenance. Newly researched SGD/MIT datasets are not in these weights.
 ## Native JSON evaluation
+Same original 1,444 SNIPS-derived source/schema requests and raw greedy protocol as the preceding release:
+|Metric|Main checkpoint|Previous native JSON release|
+|---|---:|---:|
+|Entire request correct|**83.03%**|82.48%|
+|Field accuracy|**91.36%**|91.31%|
+|Extracted-value F1|**90.40%**|90.48%|
+|Extracted-value precision|**90.53%**|90.55%|
+|Extracted-value recall|**90.26%**|90.41%|
+|Missing-field accuracy|**92.79%**|92.46%|
+|Schema-valid JSON|**99.79%**|99.79%|
+Complete correctness is 1,199/1,444 versus 1,191/1,444: 27 improved requests and 19 regressions. Unsupported values decrease 30→27. Canonical string-array views score 1,163/1,388 (83.79%) with 100% schema validity; renamed-key 21/28, nullable 15/28. The separate 54-example playground scores 44/54 (81.48%), versus 42/54; all 54 are schema-valid. The controlled new two/three-value development panel scores 14/24, versus 0/24. These are observed research panels, not fresh blind benchmarks or a public leaderboard.
+The main checkpoint was owner-selected despite strict automatic retention checks not passing. Playlist/weather exactness declines on the larger panel; aggregate precision/recall/F1 decline slightly while exact/field/missing correctness improve. The new 48-pattern panel has only 75% schema validity. Full tradeoffs and failures are documented.
+![Complete-request correctness on separate observed panels](complete-request-exact.svg)
+## Limitations
+The model remains a focused SNIPS-domain extraction research model. Valid JSON does not prove correct values. It can invent, shorten, omit or misassign names, confuse venue/movie or state/country, and mishandle multiple values or nullable fields. For example, Bergen and Oslo can generate Bergen, Orange and Lake. Broader domains, long documents, corrections, arbitrary schemas, multilingual/chat/coding quality, calibration and prompt-injection robustness are unverified. `success` only means structure-valid natural completion. No blanket reliability/nonregression claim is made.
+## Main checkpoint identity and package
+Selected BF16 weight SHA-256:
+`655fb5cfba009163cd8a7f7fe7b170b998bb452c217b1b473cfe10b33de0ccd1`
+The package includes standard model.safetensors(351,242,342 bytes), config, unchanged tokenizer, source archive, examples, evaluation records and provenance. Source/archive/weight checksums are verified by the runner. The standalone verification receipt records raw-output and stop parity across 54 local fixtures; it is not an independent blind accuracy result. Historical revisions are retained for reproducibility. Source uses Apache-2.0; upstream terms apply separately; no trained-weight license is assigned by this update.

RUN_GUIDE.md CHANGED Viewed

@@ -24,7 +24,7 @@ with raw `text`, parsed `value`, structural success, stop reason, token count an
 example exact correctness when a reference exists. The reference is compared
 only after generation and never enters the model input. Schema validity does
 not certify extracted values. Some included examples intentionally expose
-remaining errors. The current local diagnostic is 42/54, not 100% correctness.
 Example source: `Play the track Blue Lantern by Mira Sol.`
@@ -60,8 +60,11 @@ python native_bundle/inference.py --weights "$PWD/model.safetensors" --example 2
 ```
 Selected weights:
-`656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4`.
 The current source module is inference_native/native_json.py. The old labeled
 adapter, if present in the historical source archive, is not used by this
 entry point. Optimizer state and the original training workspace are not needed
 for inference. See the technical report for lineage, metrics and limitations.

 example exact correctness when a reference exists. The reference is compared
 only after generation and never enters the model input. Schema validity does
 not certify extracted values. Some included examples intentionally expose
+remaining errors. The current local diagnostic is 44/54, not 100% correctness.
 Example source: `Play the track Blue Lantern by Mira Sol.`
 ```
 Selected weights:
+`655fb5cfba009163cd8a7f7fe7b170b998bb452c217b1b473cfe10b33de0ccd1`.
 The current source module is inference_native/native_json.py. The old labeled
 adapter, if present in the historical source archive, is not used by this
 entry point. Optimizer state and the original training workspace are not needed
 for inference. See the technical report for lineage, metrics and limitations.
+The main checkpoint includes an additional cardinality/replay extraction SFT stage. The fixed inference instruction/schema/one-call native JSON protocol is unchanged; no frozen teacher is needed after download. Use example 35 to try two artists. Exact identities/history are in checkpoint_provenance.json.

SHA256SUMS CHANGED Viewed

@@ -1,20 +1,20 @@
-983da8cdd3053471db6ffe7abefe1d7ca57403be8369d3f610a18d39be9ac9ec  CITATION.cff
-a84124bd31b3191d5b1b974285361b65ac47659f3ae29adf52694c3e5ccb9a61  INFERENCE_VERIFICATION.json
-d051b0268853461b88a3ba688987e59cfa8daedab167514f39286f7c61076bac  JSONLITE_RESULTS.json
 c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4  LICENSE
-34c9fca80d9deceb4e88f7e4ea2230da468b3b030db42b1aa3b93e38dfe6fa54  PACKAGING_VERIFICATION.json
-0d97ee6e157bc68006bd75b7f98ef3313f529f63f0bb0f840fb8da6bb9b265fe  README.md
-679252594be431825d9a72ffd0a190821a6296c05fcf57b85db1d4b95cda87b9  RUN_GUIDE.md
-51746abc1e814b667fb69ee6853add348c88cdcfcd16cc94c110426759694962  SOURCE_NOTICE.md
-fa833ca4380ff47d177578067e00a186c8ce4c2bb17b0a352122ccc5437e16f5  TECHNICAL_REPORT.md
 4e9b202c1a70b476d74a4fe9b9a55e9756195a8b5bd3fc63558875a89487a91d  THIRD_PARTY_NOTICES.md
-5ecf9a8850aef6cc75add6bacda880648e72de1b38089d585e237bbce828c2c3  checkpoint_metadata.json
-21470fa38a01c8ea13886882db85d21910f9b4413e1c57a25130e5abb053b463  checkpoint_provenance.json
-83bd7d84e67ff1177057577de68fb0d3daa9afa6d0eda444b28f60ad3e2c0853  complete-request-exact.svg
-23684c5658428e3b05e3c5cec4c71163e39fee67cba8bd7b4fbde100397b1ebb  config.json
-cff12098b59ac0323a10cc6d60c6236ae9f3161b7f03277b2101607749292626  inference.py
-656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4  model.safetensors
-4ea6280d2e60da34e59757a4d6ab95aaf7811050c68c047cf5b98d2f001e9b78  native_bundle.zip
 641fe860d6fd24547fe4d15bd852190f00ddec31bd74eb7ca9eeae74031c2f0c  pretraining-training-loss.png
 c7b1a3c0efb09ef3b5416a24072e48b25e2cc1f122fb5ed638c0f95b72594af9  requirements.txt
 e095f0413a11fa91fbb58f37d00e4c6bcda59beb89b28f6f680db1436a7e8891  tokenizer.json

+dfe55d5efa7268fa3e8f2997dcb148e0145db80960cd460a242bacc81cef61a0  CITATION.cff
+980695fb234fa03113e27bad3ec2a87416168e19f4beaad98bbcad2da920f641  INFERENCE_VERIFICATION.json
+38eaa196a9bcbbb363d207b46a3c42fcc589687b7b9980e7eff218a1a8cf8bdb  JSONLITE_RESULTS.json
 c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4  LICENSE
+30291e6ad10c7f9a79108f327fc26e68d52a3d579df36f2ac6d8dacc2316e9bf  PACKAGING_VERIFICATION.json
+6586556e383ff93a3ceb4f1e1630be5dab4b73f9f3a69658ba1327416251af4a  README.md
+824603c7bd790050ad56e65102c366ede53fcf461b62ac2c6474a2a75a0fc9a5  RUN_GUIDE.md
+d5de06e61d0d6bd7245e515f41f7d3ca2d0716acc69d460f676b86f449fd078c  SOURCE_NOTICE.md
+974f3a2d6837c08d989a64ba942009c75a5d2015d32c6783c679ba6e8b47f907  TECHNICAL_REPORT.md
 4e9b202c1a70b476d74a4fe9b9a55e9756195a8b5bd3fc63558875a89487a91d  THIRD_PARTY_NOTICES.md
+49d177fb4606673f789fbf06c694643c44504672a9f0c3f75f7eae776cb4590c  checkpoint_metadata.json
+09494951de47b0cb0e2d2b683f27cf07139475b05a3fcad738bfc216ab4bc6a8  checkpoint_provenance.json
+ac810f02fd0428acc76123c281c746c70276fcee6a01b1c7a4e08e06db8097e2  complete-request-exact.svg
+4fd845ef735a67b567f0d46e123994c6dbaa18715c4b2550192720f54a280033  config.json
+2010f51d840d86ddae62fb65ac4eaaa03c90ca12f7272fe70d949f494b74d728  inference.py
+655fb5cfba009163cd8a7f7fe7b170b998bb452c217b1b473cfe10b33de0ccd1  model.safetensors
+a1b63ed6286d6c4f8bc4f7526f4a11fc46f936830daaac22224bfa0c4892755b  native_bundle.zip
 641fe860d6fd24547fe4d15bd852190f00ddec31bd74eb7ca9eeae74031c2f0c  pretraining-training-loss.png
 c7b1a3c0efb09ef3b5416a24072e48b25e2cc1f122fb5ed638c0f95b72594af9  requirements.txt
 e095f0413a11fa91fbb58f37d00e4c6bcda59beb89b28f6f680db1436a7e8891  tokenizer.json

SOURCE_NOTICE.md CHANGED Viewed

@@ -1,7 +1,7 @@
 # Source and data notice
 Current inherited weights: Stage A/132005 → SFT-065/469 → SFT-068/379 →
-SFT-093/160 → SFT-100/469. The final stage directly supervises native JSON.
 Other research trials and unused helper code are not weight ancestors.
 Author-controlled source uses Apache-2.0. Upstream datasets/dependencies retain
 their notices; no trained-weight license is assigned by this source update.
@@ -10,15 +10,19 @@ GitHub excludes weight and optimizer binaries; the HF companion has the weights.
 Pretraining uses the pinned educational web/wiki/Cosmopedia/Python mixture.
 Grounded QA uses selected SQuAD2/Dolly and checked authored fictional facts.
 SNIPS original CC0 chunk annotations supply exact extraction and native JSON
-labels. The 15,000 curated JSON records are source/schema/target views of12,115
-TRAIN requests, not15,000distinct documents. Development/test are source-disjoint
 from TRAIN within this curated dataset, but have previous research exposure.
 Pretraining decontamination and perfectly correct semantic golds are not claimed.
 Raw pretraining pools, packed arrays, optimizer state, credentials, SSH keys and
 private host/network receipts are excluded. Published metadata replaces private
 workspace paths with [workspace]; original and published hashes are separated.
-54actual local fixtures, raw predictions and held-out evidence are retained.
-The canonical held-out1,388and schema-stress56requests are included as JSONL in
 evaluation/native-json-heldout.jsonl and the prepared native data directory.
 See THIRD_PARTY_NOTICES.md, docs/TECHNICAL_REPORT.md and provenance/LINEAGE.json.

 # Source and data notice
 Current inherited weights: Stage A/132005 → SFT-065/469 → SFT-068/379 →
+SFT-093/160 → SFT-100/469 → additional extraction SFT (main checkpoint). The final stage directly supervises native JSON.
 Other research trials and unused helper code are not weight ancestors.
 Author-controlled source uses Apache-2.0. Upstream datasets/dependencies retain
 their notices; no trained-weight license is assigned by this source update.
 Pretraining uses the pinned educational web/wiki/Cosmopedia/Python mixture.
 Grounded QA uses selected SQuAD2/Dolly and checked authored fictional facts.
 SNIPS original CC0 chunk annotations supply exact extraction and native JSON
+labels. The 15,000 curated JSON records are source/schema/target views of 12,115
+TRAIN requests, not 15,000 distinct documents. Development/test are source-disjoint
 from TRAIN within this curated dataset, but have previous research exposure.
 Pretraining decontamination and perfectly correct semantic golds are not claimed.
 Raw pretraining pools, packed arrays, optimizer state, credentials, SSH keys and
 private host/network receipts are excluded. Published metadata replaces private
 workspace paths with [workspace]; original and published hashes are separated.
+54 actual local fixtures, raw predictions and held-out evidence are retained.
+The canonical held-out 1,388 and schema-stress 56 requests are included as JSONL in
 evaluation/native-json-heldout.jsonl and the prepared native data directory.
 See THIRD_PARTY_NOTICES.md, docs/TECHNICAL_REPORT.md and provenance/LINEAGE.json.
+The additional stage uses original SNIPS TRAIN replay plus authored cardinality
+views, with frozen-parent training-only KL. Main inherits 2,048 presentations, not
+the entire 3,072-presentation completed experiment. See the report for exact counts.

TECHNICAL_REPORT.md CHANGED Viewed

@@ -4,17 +4,17 @@
 ## 1. Current artifact and behavior
-The current weights directly generate JSON from source text and a supplied JSON Schema. A single native autoregressive request emits the entire object. There is no fieldwise extraction-to-JSON assembler, answer repair, constrained decoder or gold-conditioned filling in the current inference path. Parsing and schema validation happen only after generation; the raw generated string is retained even when incorrect or invalid.
-Current artifact: SFT-100/update 469, BF16 `model.safetensors`, 351,242,342 bytes.
-SHA-256: `656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4`.
 This is a custom native MLX architecture, not a Transformers/MLX-LM auto-loader export.
-This report supersedes the earlier fieldwise SFT-093 release. Its historical report is retained as `docs/HISTORICAL_SFT093_REPORT.md` inside the source archive; those adapter-assisted scores are not current native JSON scores. No release-readiness or arbitrary-schema reliability claim is made.
 ## 2. Exact inherited lineage
-`Stage A/132005 → SFT-065/469 → SFT-068/379 → SFT-093/160 → SFT-100/469`
 Only those selected checkpoints are inherited. Other trials and failed RL/JSON experiments are research history, not hidden additional ancestors. The architecture and frozen tokenizer remain unchanged. Each SFT stage resets its local counters and AdamW optimizer while continuing the inherited neural weights.
@@ -25,7 +25,8 @@ Only those selected checkpoints are inherited. Other trials and failed RL/JSON e
 | SFT-068 requested-field extraction | 379 | 12,115 | 1,842,593 |
 | SFT-093 focused extraction/replay | 160 | 5,120 | 706,399 |
 | SFT-100 native JSON | 469 | 15,000 | 4,634,126 |
-| Inherited post-training total | 1,477 updates | 47,243 | 10,280,449 |
 Presentation counts include repeated source views. Serialized positions include masked prompt tokens; they are not supervised-output-token counts. The earlier 150M model weights did not initialize this model; the earlier frozen OpenSML tokenizer was reused.
@@ -66,10 +67,10 @@ The model uses the project's frozen 32K byte-level BPE tokenizer at `reference/o
 | Tokenizer artifact | SHA-256 |
 | --- | --- |
-| Tokenizer manifest / contract fingerprint | 1194292c6d906ac19e6f01c6ad2b42825143fad426a22aa8856d903776701ecb |
-| tokenizer.json | e095f0413a11fa91fbb58f37d00e4c6bcda59beb89b28f6f680db1436a7e8891 |
-| tokenizer_config.json | 619f500553367eb76f5be00a98625130ec857b9c367561334984096c029ba2a9 |
-| tokenizer corpus.json | e4c47b7b6aba08f948de7c2c7cb94c58dbd09dde2ded6ac185c60856ee686d27 |
 Structural token IDs:
@@ -92,11 +93,11 @@ Source weights below are intended processed-token sampling weights, not measured
 | Source | Sampling weight (%) | Repository / configuration | Pinned revision | Additional selection |
 | --- | --- | --- | --- | --- |
-| FineWeb-Edu | 50 | HuggingFaceTB/smollm-corpus / fineweb-edu-dedup | 3ba9d605774198c5868892d7a8deda78031a781f | Deduplicated educational subset |
-| DCLM-Edu | 18 | HuggingFaceTB/dclm-edu | dbad8ad71224482740cd9c9d353591adbf62fe04 | edu_int_score≥3 |
-| FineWiki English | 13 | HuggingFaceFW/finewiki / en | 8bd13e72e6a002407649b3e898535f42ceb1aeb9 | English |
-| Cosmopedia v2 | 9 | HuggingFaceTB/smollm-corpus / cosmopedia-v2 | 3ba9d605774198c5868892d7a8deda78031a781f | Format contains textbook/tutorial/blog/educational |
-| Python text mirror | 10 | Avelina/python-edu-cleaned / default | 3fedf55e02e72dd1fd68963f991f5e91adcb0f8d | int_score≥4 |
 Public source cards: [SmolLM corpus](https://huggingface.co/datasets/HuggingFaceTB/smollm-corpus/blob/3ba9d605774198c5868892d7a8deda78031a781f/README.md), [DCLM-Edu](https://huggingface.co/datasets/HuggingFaceTB/dclm-edu/blob/dbad8ad71224482740cd9c9d353591adbf62fe04/README.md), [FineWiki](https://huggingface.co/datasets/HuggingFaceFW/finewiki/blob/8bd13e72e6a002407649b3e898535f42ceb1aeb9/README.md). The Python-Edu upstream subset identifies files; this run used the configured text mirror. The mirror's pinned README could not be retrieved during report preparation, so its provenance statement here comes from the local corpus specification, not a newly completed external audit.
@@ -108,7 +109,7 @@ Recorded stream settings are a 256-document accepted-document shuffle buffer, bo
 ## 6. Pretraining execution, optimization and validation
-Pretraining runs on the recorded five-Mac Apple Silicon cluster. Mac1's current local hardware was independently checked during report preparation as an Apple M5 Ultra with 96 GB unified memory. Other worker descriptions below are historical cluster records, not newly queried hardware inventories.
 | Mac | Recorded chip | RAM | Microbatch | MLX cache cap |
 | --- | --- | --- | --- | --- |
@@ -120,9 +121,9 @@ Pretraining runs on the recorded five-Mac Apple Silicon cluster. Mac1's current
 The physical ring is 1→2→3→5→4→1. Four accumulation microsteps and 2,048-token sequences produce `(7+3+2+2+2)×4×2048=131072` positions/update. Unequal-rank gradients are weighted by token counts. Early batch transitions mean the lifetime token count is not simply `132005×131072`.
-- First 125 updates: microbatches [4, 4, 2, 2, 2], 114,688 positions/update, 14,336,000 tokens.
-- From 125: microbatches [6, 4, 2, 2, 2], 131,072 positions/update.
-- From 963 / 124,174,336 tokens: [7, 3, 2, 2, 2], same global batch; per-Mac cache caps recorded.
 - Stage A continuation starts from 113313 / 14,850,113,536 tokens, retaining weights, FP32 optimizer state, source cursors, counters and controller history.
 - Selected base: 132005 / 17,300,111,360 tokens.
@@ -149,7 +150,7 @@ The nominal 3e-4 schedule is not the effective late-run learning rate: the retai
 Validation and qualitative samples occur at 50M-token boundaries. The fixed validation panel has 32 batches/source, batch 2, five sources, context 2,048: nominally 655,360 positions/event. Aggregate loss is source-mixture weighted. Validation fingerprint: `4ddd461c131d4d27d07bb07f5aab3fcb5a00d4b1202066e3d9b6ee6b3129087b`. Selected-base validation loss is **2.5563359551876785**, perplexity **12.8885**. This measures document prediction, not extraction or chat reliability.
-Regular saves occur every 100M tokens, plus best-validation and clean-stop saves. Exact checkpoint metadata is authoritative. The initial October 2 technical report's early observations, 20B target and old schedule describe history; they are superseded here for this selected artifact. Contemporary qualitative logs often showed roughly 41k tokens/s, but no reproducible end-to-end cluster performance/energy benchmark is established by this report. The separate 42.7k profiling candidate is not claimed as the configuration that trained this checkpoint.
 ### Training run
@@ -160,155 +161,109 @@ The 175.6M-parameter base was trained from random initialization using next-toke
 *Training loss and held-out validation during pretraining. Left: raw training loss and exponential moving average. Right: held-out validation loss.*
-## 7. Inherited supervised stages and objective
-SFT-065 starts directly from Stage A/132005. Its prepared 25,000-record grounded-QA corpus contains 12,000 natural SQuAD 2.0/Dolly QA records, 5,000 contrasting questions over shared evidence, 3,000 answerability/missing-information cases, 3,000 multipart questions and 2,000 corrections. Unsupported answers target `Not stated.`. The whole run reached 782 updates; only update 469 and its 15,008 presentations are inherited.
-SFT-068 uses 12,115 curated original SNIPS TRAIN requests across seven intent domains. Each supplies requested field names/definitions and source text. Assistant targets are labeled exact source spans, with `NOT_STATED` for absent fields. It completes 379 updates. Original joint-labeled accuracy is a different protocol from native JSON and is not attributed to the current weights.
-SFT-093 uses 2,560 original SNIPS TRAIN queries (1,920 accepted by a TRAIN-only parent-output verification bank and 640 additional queries). Each contributes a labeled replay view and a focused single-field view, yielding 5,120 presentations. Every batch has 16 replay and 16 focused examples. This stage completes 160 updates. It was not a native JSON target stage; its later software-assisted JSON route is historical.
-All four SFT stages update full weights on Mac1 with assistant-output-only, conversation-mean cross entropy. User/schema/source tokens and role headers are masked; assistant content and turn/document stops are supervised. BF16 model computation uses FP32 master AdamW weights/moments and FP32 gradient accumulation. No LoRA, frozen-layer training, RL, added entity head or weight merge is part of this lineage.
-| Setting | SFT-065 | SFT-068 | SFT-093 | SFT-100 |
-| --- | --- | --- | --- | --- |
-| Batch / microbatch | 32 / 2 | 32 / 2 | 32 / 2 | 32 / 2 |
-| Planned updates | 782 | 379 | 160 | 469 |
-| Inherited selected update | 469 | 379 | 160 | 469 |
-| Peak / final LR | 1e-5 / 2e-6 | 5e-6 / 1e-6 | 5e-6 / 1e-6 | 5e-6 / 1e-6 |
-| Warmup updates | 20 | 20 | 12 | 20 |
-| Weight decay / clip | 0.01 / 1 | 0.01 / 1 | 0.01 / 1 | 0.01 / 1 |
-| Adam betas / epsilon | (0.9, 0.95) / 1e-8 | Same | Same | Same |
-| LR decay | Cosine | Cosine | Cosine | Cosine |
-## 8. SFT-100: source text + schema → JSON targets
-The final one-pass continuation starts from preserved SFT-093/update 160 with a fresh optimizer. Every assistant target is an actual JSON object. No labeled-line replay target is mixed into this stage.
-| JSON training view | Records |
-| --- | ---: |
-| Canonical string-array schemas | 12,115 |
-| Renamed property keys with field descriptions | 1,443 |
-| Nullable-string schema views | 1,442 |
-| Total presentations | 15,000 |
-| Distinct source requests | 12,115 |
-The additional views reuse TRAIN-only sources. Original SNIPS annotation chunks at revision `b86ac7f1577868c42158d0dec77db50956046696` provide exact values and offsets. All targets are reconstructed from those annotations, compared with the source text and validated against the supplied schema. This is annotation fidelity, not a claim that every upstream annotation is independently perfect semantic truth.
-Schemas are closed objects with all properties required. Arrays contain all annotated source spans in source order, or `[]` when absent. Nullable strings are used only when the gold has at most one value, with `null` for absence. Multi-value fields remain arrays. Spelling and case are copied exactly; numbers/dates are not normalized. TRAIN alias keys use `out_`; DEV/TEST use `selected_`. Gold key maps are curation/evaluation metadata only and never enter inference.
-The seven domains are AddToPlaylist, BookRestaurant, GetWeather, PlayMusic, RateBook, SearchCreativeWork and SearchScreeningEvent; there are 39 trained canonical field definitions. TRAIN/DEV/TEST source-text groups do not overlap within this prepared dataset. The frozen development/test panels had previous research exposure; pretraining decontamination and perfect semantic golds are not claimed.
-Packed TRAIN has 4,634,126 serialized positions, max prompt length 425 and max target length 75. No record is truncated. Scheduled DEV checks at 0, 100, 200, 300 and 469 use 252 requests: 196 canonical, 28 alias and 28 nullable views of 98 source requests. TEST contains 1,444 requests: 1,388 canonical plus 28 alias and 28 nullable views of 694 distinct sources. Packed arrays are not in the source repository; JSONL records and the preparation/masking code are included. The original data/checkpoint manifests record workspace files (including excluded packed arrays/optimizer binaries); use provenance/SNAPSHOT_MANIFEST.json for the actual distributed file inventory.
-The local preflight checked target/schema/annotation agreement, evidence offsets, tokenization and assistant masks, split separation and frozen hashes. Eleven adversarial scorer/retention tests passed. GPU forward/backward accumulation was checked without optimizer updates or new checkpoints. Batch 32 uses actual microbatch 2, padding each microbatch to a multiple of 128 with no unused 64-slot fake samples. Training completed 469 updates, 15,000 presentations and 4,634,126 positions.
-## 9. Native inference contract
-Input is source text plus the supplied JSON Schema, serialized in a fixed extraction instruction followed by compact `{"source_text": ..., "schema": ...}`. Native role IDs construct a user turn and the assistant-generation header. The entire source/schema prompt reaches the model together. There is no canonical-field compiler that extracts one field at a time.
-The portable runner accepts one to eight required properties in a closed flat object: strings, nullable strings (`["string","null"]`) or arrays with `items: {"type":"string"}`. `$schema`, when present, is draft 2020-12. Nested records, numbers, booleans, enums, references and advanced constraints are outside this verified runner contract. Unknown names/descriptions are passed directly to the neural model, but reliable unfamiliar-domain semantics are not established.
-Default generation is greedy BF16 native MLX, temperature 0, repetition penalty 1, max 256 output tokens. Prompt plus output budget must fit 2,048 tokens; overflow rejects without truncation. Stop IDs are turn-end 3 or document-end 1; unexpected structural tokens or token exhaustion fail complete correctness. JSON object key order is irrelevant; array order and exact source strings matter.
-The default CLI writes raw model text. `--json` returns a result record with raw `text`, parsed `value`, schema/stop status and diagnostics. Parsing is strictly post-generation; duplicate keys and nonfinite JSON numbers reject. No generated value is filled, normalized, repaired or reserialized to create the model's raw output. The example reference stays caller-side and is compared only after inference.
-```bash
-python inference.py --weights model.safetensors --example 29
-python inference.py --weights model.safetensors --text-file request.txt --schema-file schema.json
-python inference.py --weights model.safetensors --example 29 --json
-```
-Example 29 source: `Play the track Blue Lantern by Mira Sol.` The actual current native output is `{"track":["Blue Lantern"],"artist":["Mira Sol"],"album":[]}`. The schema names/describes those fields as arrays of strings. The model generates all keys and values in one request.
-## 10. Evaluation and checkpoint selection
-Selection was frozen from development before TEST generation. Rank prioritizes domain-macro complete correctness, then overall complete correctness, field accuracy, F1, schema validity, paired correctness, unsupported-value count and NLL. Eligibility requires no loss of complete-request or field correctness on any fixed domain/schema panel relative to the same-protocol raw parent at update 0. This is a limited observed-panel gate, not protection against every regression and not a comparison with the parent's old assembler.
-Complete correctness requires natural completion, strict JSON parsing, schema validity and exact agreement with every gold value. Fences, trailing prose, duplicate keys, nonfinite numbers, missing/extra keys, wrong types or values, wrong array ordering and exhausted budgets fail. Field accuracy is exact per requested field. Value precision/recall/F1 use `(field,value)` multisets. Unsupported values use literal source-substring checks and do not detect every association or boundary error.
-| Development update | Complete correct / 252 | Field accuracy | Value F1 | Schema-valid rate |
-| --- | --- | --- | --- | --- |
-| 0 | 0 (0%) | 0% | 0% | 0% |
-| 100 | 88 (34.92%) | 59.11% | 63.64% | 97.22% |
-| 200 | 162 (64.29%) | 80.30% | 81.03% | 99.60% |
-| 300 | 177 (70.24%) | 85.76% | 84.71% | 99.21% |
-| 469 selected | 200 (79.37%) | 90.40% | 89.99% | 99.60% |
-### Selected candidate held-out results: native generation
-| Metric | All 1,444 requests | Canonical 1,388 requests |
-| --- | ---: | ---: |
-| Entire request correct | 1,191 (82.48%) | 1,156 (83.29%) |
-| Field accuracy | 3,193/3,497 (91.31%) | 3,074/3,347 (91.84%) |
-| Value precision | 90.55% | 90.95% |
-| Value recall | 90.41% | 90.90% |
-| Value F1 | 90.48% | 90.92% |
-| Present-field accuracy | 1,782/1,971 (90.41%) | 1,708/1,879 (90.90%) |
-| Missing-field accuracy | 1,411/1,526 (92.46%) | 1,366/1,468 (93.05%) |
-| False abstentions | 111 | 103 |
-| Missed abstentions | 115 | 102 |
-| Unsupported values | 30 | 24 |
-| Strict JSON parse | 1,442 (99.86%) | 1,388 (100%) |
-| Schema-valid output | 1,441 (99.79%) | 1,388 (100%) |
-| Natural stops / token limits | 1,444 / 0 | 1,388 / 0 |
-| Output NLL | 0.028335 | 0.026668 |
-Both canonical requested-field views are correct for 501/694 source pairs (72.19%). Alias stress is 19/28 (67.86%); nullable stress is 16/28 (57.14%). These extra schemas reuse held-out source texts, not additional independent documents.
-| Domain, including stress views | Complete-request exact |
-| --- | ---: |
-| AddToPlaylist | 168/206 (81.55%) |
-| BookRestaurant | 156/208 (75.00%) |
-| GetWeather | 168/206 (81.55%) |
-| PlayMusic | 149/208 (71.63%) |
-| RateBook | 188/204 (92.16%) |
-| SearchCreativeWork | 197/208 (94.71%) |
-| SearchScreeningEvent | 165/204 (80.88%) |
-The automatic same-protocol raw-parent comparison completes with parent 0/1,444 exact versus candidate 1,191/1,444. The parent had not been trained to generate native JSON from this combined schema/source prompt. Its historical fieldwise-plus-assembler score is a separate protocol: 1,132/1,388 (81.56%), compared with current native canonical 1,156/1,388 (83.29%). Current canonical false abstentions and unsupported values are higher (103 and 24 versus 69 and 5), despite improved complete correctness. Do not claim universal retention or equate these different inference protocols.
-### All 54 actual playground examples
-The complete live SSE test keeps the displayed schema ordering, instruction, source and settings unchanged. Model outputs are stored and rescored verbatim; reference answers never enter a generation POST.
-| Metric / group | Result |
-| --- | ---: |
-| Whole request exact | 42/54 (77.78%) |
-| Strict parse / schema validity | 52/54 (96.30%) |
-| Field accuracy | 118/135 (87.41%) |
-| Value precision / recall / F1 | 91.03% / 84.52% / 87.65% |
-| Present / missing-field correctness | 70/82 (85.37%) / 48/53 (90.57%) |
-| False / missed abstentions | 3 / 5 |
-| Unsupported values | 2 |
-| Natural stop / token limits | 54 / 0 |
-| Original observed SNIPS views | 25/28 (89.29%) |
-| Previously authored natural diagnostics | 7/12 (58.33%) |
-| Additional schema stress views | 10/14 (71.43%) |
-These are observed adapted SNIPS and local diagnostic panels, not official SNIPS intent classification or a fresh blind general text-to-JSON benchmark. No Paraloq, SOB, JSONSchemaBench, ARC, IFEval or coding score is attributed to these exact weights. General language quality is not established by JSON schema validity. Playground output NLL was not measured through the API.
-## 11. Failure evidence and intended scope
-The retained live outputs include Oslo changed to unsupported Odense, two performers combined into one string, a second city omitted, a maximum rating omitted, missing Thai cuisine, restaurant type confused with restaurant name, incomplete spans, a duplicate-key object and malformed nullable-string JSON. The model is a focused small-model extraction research candidate; it is not established as a reliable arbitrary-schema or general web-document extractor.
-Schema validity does not verify values. Important gaps remain in exact copying, multiple values, boundaries/associations, correction/current-value reasoning, nullable fields, unfamiliar descriptions, long passages and schema complexity. General dialogue, coding, multilingual transfer, calibrated confidence, prompt-injection robustness and comprehensive safety/privacy behavior are unverified. Release readiness remains false in the research record.
-## 12. Runtime, packaging and verification
-The current portable inference package is tested independently of the training workspace on Apple Silicon macOS, Python 3.11 with stock MLX 0.32.3, NumPy 2.4.2, Tokenizers 0.22.2 and jsonschema 4.26.0. Greedy raw output text and stop reasons are compared with all 54 live playground outputs. The verification artifact records coverage and measured correctness; this is packaging parity, not a fresh blind accuracy result or full 1,444-case cross-runtime parity claim.
-Hugging Face provides canonical `model.safetensors`, config/tokenizer files, a root runner and `native_bundle.zip`; the archive includes the native model implementation, prompt/validator, tokenizer companions, examples and reproducibility records. GitHub contains source/documentation and provenance, with weights hosted separately. Optimizer binaries, packed NumPy arrays, host network logs and private workspace paths are excluded. Original/published metadata hashes are separated in lineage provenance.
-## 13. Reproduction and evidence
-The frozen source/schema/JSON records, preparation/masking code, manifests, selected training metadata, DEV evaluations, parent/candidate TEST outputs and all 54 local predictions are included. Training uses the original workspace modules and the preserved SFT-093 parent; reproducing training requires reconstructing the corresponding workspace inputs/checkpoint paths and regenerating a local preflight receipt. The downloadable inference package does not require that workspace or optimizer state. Publishing training source is not a claim that all historic pretraining raw data or discarded optimizer checkpoints are included.
-See `../provenance/LINEAGE.json`, `../results/JSONLITE_RESULTS.json`, `../reports/model-design/SFT-100_native_json_20261014/`, `../pretraining/native_json_100/`, `../examples/cases.json` and `../inference_native/INFERENCE_VERIFICATION.json` in the source archive. `scripts/verify_snapshot.py` checks the source manifest; the HF root runner checks the source archive and weight identities. Root tokenizer artifacts mirror the archive's frozen tokenizer.
-## 14. Attribution and licensing
-Source code carries the repository's Apache-2.0 license; upstream data/dependency notices retain their terms. SNIPS original entity annotations are CC0-1.0. SQuAD/Dolly and the pretraining mixture have separate source notices recorded in `THIRD_PARTY_NOTICES.md`. Source licensing does not automatically assign a trained-weight license; the owner has not assigned one in this update. No blanket commercial-use or redistribution determination is claimed.
-Hugging Face edition: archive-relative source/evidence references resolve inside `native_bundle.zip`; root `RUN_GUIDE.md` explains download and native inference.

 ## 1. Current artifact and behavior
+The main checkpoint directly generates JSON from source text and a supplied JSON Schema. A single native autoregressive request emits the entire object. There is no fieldwise extraction-to-JSON assembler, answer repair, constrained decoder or gold-conditioned filling in the current inference path. Parsing and schema validation happen only after generation; the raw generated string is retained even when incorrect or invalid.
+Current artifact: **main checkpoint**, BF16 `model.safetensors`, 351,242,342 bytes.
+SHA-256: `655fb5cfba009163cd8a7f7fe7b170b998bb452c217b1b473cfe10b33de0ccd1`.
 This is a custom native MLX architecture, not a Transformers/MLX-LM auto-loader export.
+This report supersedes the earlier fieldwise SFT-093 release. Its historical report is retained as `HISTORICAL_SFT093_REPORT.md`; those adapter-assisted scores are not current native JSON scores. No release-readiness or arbitrary-schema reliability claim is made.
 ## 2. Exact inherited lineage
+`Stage A → grounded QA → requested-field extraction → focused extraction → native JSON foundation → additional extraction SFT`
 Only those selected checkpoints are inherited. Other trials and failed RL/JSON experiments are research history, not hidden additional ancestors. The architecture and frozen tokenizer remain unchanged. Each SFT stage resets its local counters and AdamW optimizer while continuing the inherited neural weights.
 | SFT-068 requested-field extraction | 379 | 12,115 | 1,842,593 |
 | SFT-093 focused extraction/replay | 160 | 5,120 | 706,399 |
 | SFT-100 native JSON | 469 | 15,000 | 4,634,126 |
+| Additional extraction SFT | 64 selected of 96 completed | 2,048 | 600,558 |
+| Inherited post-training total | 1,541 updates | 49,291 | 10,881,007 |
 Presentation counts include repeated source views. Serialized positions include masked prompt tokens; they are not supervised-output-token counts. The earlier 150M model weights did not initialize this model; the earlier frozen OpenSML tokenizer was reused.
 | Tokenizer artifact | SHA-256 |
 | --- | --- |
+| Tokenizer manifest / contract fingerprint | 1194292 c 6 d 906 ac 19e 6 f 01 c 6 ad 2 b 42825143 fad 426 a 22 aa 8856 d 903776701ecb |
+| tokenizer.json | e 095 f 0413 a 11 fa 91 fbb 58 f 37 d 00e 4 c 6 bcda 59 beb 89 b 28 f 6 f 680 db 1436 a 7e 8891 |
+| tokenizer_config.json | 619 f 500553367eb 76 f 5 be 00 a 98625130ec 857 b 9 c 367561334984096 c 029 ba 2 a 9 |
+| tokenizer corpus.json | e 4 c 47 b 7 b 6 aba 08 f 948 de 7 c 2 c 7 cb 94 c 58 dbd 09 dde 2 ded 6 ac 185 c 60856ee 686 d 27 |
 Structural token IDs:
 | Source | Sampling weight (%) | Repository / configuration | Pinned revision | Additional selection |
 | --- | --- | --- | --- | --- |
+| FineWeb-Edu | 50 | HuggingFaceTB/smollm-corpus / fineweb-edu-dedup | 3 ba 9 d 605774198 c 5868892 d 7 a 8 deda 78031 a 781 f | Deduplicated educational subset |
+| DCLM-Edu | 18 | HuggingFaceTB/dclm-edu | dbad 8 ad 71224482740 cd 9 c 9 d 353591 adbf 62 fe 04 | edu_int_score≥3 |
+| FineWiki English | 13 | HuggingFaceFW/finewiki / en | 8 bd 13e 72e 6 a 002407649 b 3e 898535 f 42 ceb 1 aeb 9 | English |
+| Cosmopedia v2 | 9 | HuggingFaceTB/smollm-corpus / cosmopedia-v2 | 3 ba 9 d 605774198 c 5868892 d 7 a 8 deda 78031 a 781 f | Format contains textbook/tutorial/blog/educational |
+| Python text mirror | 10 | Avelina/python-edu-cleaned / default | 3 fedf 55e 02e 72 dd 1 fd 68963 f 991 f 5e 91 adcb 0 f 8 d | int_score≥4 |
 Public source cards: [SmolLM corpus](https://huggingface.co/datasets/HuggingFaceTB/smollm-corpus/blob/3ba9d605774198c5868892d7a8deda78031a781f/README.md), [DCLM-Edu](https://huggingface.co/datasets/HuggingFaceTB/dclm-edu/blob/dbad8ad71224482740cd9c9d353591adbf62fe04/README.md), [FineWiki](https://huggingface.co/datasets/HuggingFaceFW/finewiki/blob/8bd13e72e6a002407649b3e898535f42ceb1aeb9/README.md). The Python-Edu upstream subset identifies files; this run used the configured text mirror. The mirror's pinned README could not be retrieved during report preparation, so its provenance statement here comes from the local corpus specification, not a newly completed external audit.
 ## 6. Pretraining execution, optimization and validation
+Pretraining runs on the recorded five-Mac Apple Silicon cluster. Mac 1's current local hardware was independently checked during report preparation as an Apple M5 Ultra with 96 GB unified memory. Other worker descriptions below are historical cluster records, not newly queried hardware inventories.
 | Mac | Recorded chip | RAM | Microbatch | MLX cache cap |
 | --- | --- | --- | --- | --- |
 The physical ring is 1→2→3→5→4→1. Four accumulation microsteps and 2,048-token sequences produce `(7+3+2+2+2)×4×2048=131072` positions/update. Unequal-rank gradients are weighted by token counts. Early batch transitions mean the lifetime token count is not simply `132005×131072`.
+- First 125 updates: microbatches [4,4,2,2,2], 114,688 positions/update, 14,336,000 tokens.
+- From 125: microbatches [6,4,2,2,2], 131,072 positions/update.
+- From 963 / 124,174,336 tokens: [7,3,2,2,2], same global batch; per-Mac cache caps recorded.
 - Stage A continuation starts from 113313 / 14,850,113,536 tokens, retaining weights, FP32 optimizer state, source cursors, counters and controller history.
 - Selected base: 132005 / 17,300,111,360 tokens.
 Validation and qualitative samples occur at 50M-token boundaries. The fixed validation panel has 32 batches/source, batch 2, five sources, context 2,048: nominally 655,360 positions/event. Aggregate loss is source-mixture weighted. Validation fingerprint: `4ddd461c131d4d27d07bb07f5aab3fcb5a00d4b1202066e3d9b6ee6b3129087b`. Selected-base validation loss is **2.5563359551876785**, perplexity **12.8885**. This measures document prediction, not extraction or chat reliability.
+Regular saves occur every 100M tokens, plus best-validation and clean-stop saves. Exact checkpoint metadata is authoritative. The initial October 2 technical report's early observations, 20B target and old schedule describe history; they are superseded here for this selected artifact. Contemporary qualitative logs often showed roughly 41 k tokens/s, but no reproducible end-to-end cluster performance/energy benchmark is established by this report. The separate 42.7 k profiling candidate is not claimed as the configuration that trained this checkpoint.
 ### Training run
 *Training loss and held-out validation during pretraining. Left: raw training loss and exponential moving average. Right: held-out validation loss.*
+## 7. Supervised foundation and additional extraction SFT
+The four inherited foundation stages remain unchanged: grounded evidence QA (15,008 presentations through its selected checkpoint), requested-field SNIPS extraction (12,115), focused single-field extraction/replay(5,120), then directly supervised native JSON (15,000). Their original source and checkpoint records remain in provenance. Each stage uses full model weights, fresh AdamW, BF16 computation with FP32 master weights/moments/gradient accumulation, and assistant-output-only supervision. No RL, LoRA or added extraction head is in this lineage.
+The native JSON foundation has 12,115 distinct SNIPS TRAIN requests, with 12,115 canonical string-array schemas, 1,443 renamed-key views and 1,442 nullable-string views. Source text, field definitions and the complete schema are input; complete JSON and stop tokens are the assistant target. Original CC0 SNIPS chunk annotations provide source spans. Metadata fidelity and schema validation do not establish perfect semantic annotation. Frozen tokenizer and architecture are unchanged.
+### Additional extraction SFT
+This completed continuation starts from the native JSON foundation. It adds matched zero/one/two/three-value supervision for artists and cities, including zero-primary-value cases where an album or country is explicitly present. It retains original JSON TRAIN replay and adds frozen-parent distribution distillation on replay assistant tokens. Every model batch has 24 replay and 8 new views. Public checkpoint name: **main**; the neural tensor is `model.safetensors`.
+| Setting | Value |
+|---|---|
+| Prepared corpus |3,072 presentations: 2,304 original TRAIN replay +768 authored schema views|
+| Distinct authored source requests in prepared corpus |256|
+| Prepared replay variants |1,344 canonical, 480 renamed, 480 nullable|
+| Completed experiment |96 optimizer updates, one pass, 900,453 serialized positions|
+| Inherited selected prefix |2,048 presentations, 600,558 positions, 64 updates|
+| Inherited prefix mixture |1,536 replay views from 1,482 source texts +512 authored views from 250 source texts|
+| Model adaptation |Full weights; Mac 1 only; fresh AdamW|
+| Batch / microbatch |32 /2|
+| Peak / final schedule LR |2e-6 /2e-7 over 96 updates|
+| Warmup |6 updates; cosine afterward|
+| Betas / epsilon |(0.9,0.95) /1e-8|
+| Weight decay / clipping |0.01 /1.0|
+| Seed |20261119|
+The main checkpoint inherits only the selected prefix; updates after it are not part of these weights. The endpoint's actual learning rate is recorded by the cosine schedule, not the planned final LR. Presentation counts include multiple schemas for the same text and do not mean distinct user queries.
+For every row, CE is averaged over supervised assistant tokens within each sequence, then averaged across sequences. On original TRAIN replay add 1.0×KL(frozen foundation || student), using the full 32,000-token distribution, temperature 1 and teacher-forced ground-truth histories. Only assistant JSON and turn/document-stop positions participate. The reference network is frozen/stop-gradient; new authored rows use CE alone. This regularizer is training-only: no teacher is needed for inference, and it does not guarantee retention or repair parent errors.
+The new views use literal original SNIPS TRAIN values with split-disjoint authored value pools/phrases, excluding observed evaluation source texts and holding new authored DEV/TEST values out of replay. Source text, exact offsets, schema/gold agreement, masks, token budgets and hashes were checked. Twenty-eight tests, GPU accumulation checks, frozen-reference checks and a temporary optimizer/checkpoint-resume smoke passed. Authored examples are programmatic controlled supervision, not independently collected user traffic. SGD, MIT Movie and other newly researched datasets are **not** in this checkpoint.
+## 8. Prompt, schema and native inference
+Supply source text and a flat closed JSON Schema with 1–8 required fields. Supported properties are strings, nullable strings and arrays of strings. Every required key is returned; absent arrays target [] and absent nullable strings target null. Copy case/spelling and complete values; preserve source order. Nested schemas, numbers/booleans, enums and advanced constraints are outside the verified runner contract. The supplied runner rejects overlong inputs without truncation.
+The fixed instruction is followed by compact `{"source_text":...,"schema":...}` and native SFT role tokens. It issues exactly one autoregressive whole-object generation with temperature 0, repetition penalty 1 and at most 256 output tokens. No fieldwise assembly, repair, schema-constrained decoder, normalization, reference answer, frozen teacher or secondary model is used at inference. Parsing/schema checks only annotate the raw generated text afterward.
+## 9. Evaluation protocol and selection
+The continuation checks 404 fixed development requests plus 54 local playground fixtures at its scheduled checkpoints. DEV contains 252 original native JSON controls, 64 previously observed combined-request controls, 8 earlier single-request probes, 32 earlier authored list controls and 48 new cardinality controls. Metrics include exact requests/fields, precision/recall/F1, present/missing fields, false/missed abstentions, unsupported values, strict parsing, schema validity, stopping/budget failures, paired requests and cardinality-group correctness. These panels have prior research exposure; no new blind/general extraction benchmark claim is made.
+The main checkpoint is an owner-selected research tradeoff. Strict automatic retention gates did not pass: some domain/variant unsupported values and nullable false abstentions increased. The run's automatic best pointer remains its starting baseline; the published main is explicitly the selected trained checkpoint, not a claim that the automatic gate accepted it. Larger original TEST comparison was performed after the candidate was nominated from development/playground results. Subsequent owner release choice observed that comparison. The original 1,444 panel also has earlier research exposure.
+### Main checkpoint versus previous native JSON release
+Identical original 1,444 source/schema requests, tokenizer, native prompt, raw greedy decoding, 256-token budget and scorer were used. Previous release outputs were identity-checked and rescored from their saved cache. The main checkpoint generated every request, with no dropped cases or token limits. Complete correctness requires natural stopping, strict JSON/schema validity and exact field/value/order agreement. Value scores use(field, value) multisets; literal unsupported-value checks do not detect every wrong association.
+|Metric|Previous native JSON release|Main checkpoint|
+|---|---:|---:|
+|Entire request correct|82.4792%|83.0332%|
+|Field accuracy|91.3068%|91.3640%|
+|Extracted-value F1|90.4798%|90.3963%|
+|Extracted-value precision|90.5488%|90.5344%|
+|Extracted-value recall|90.4110%|90.2588%|
+|Missing-field accuracy|92.4640%|92.7916%|
+|Schema-valid JSON|99.7922%|99.7922%|
+Complete requests increase 1,191→1,199/1,444 (+8).27 previously wrong requests become correct and 19 previously correct become wrong; those 19 schema views cover 18 unique source texts. Paired source correctness increases 501→512/694. Unsupported values 30→27; false abstentions 111→107; missed missing fields 115→110. Strict parsing remains 1,442/1,444 and schema validity 1,441/1,444. These modest aggregate gains coexist with small F1/precision/recall decreases.
+|Domain|Previous correct|Main correct|Requests|
+|---|---:|---:|---:|
+|AddToPlaylist|168|165|206|
+|BookRestaurant|156|162|208|
+|GetWeather|168|164|206|
+|PlayMusic|149|153|208|
+|RateBook|188|193|204|
+|SearchCreativeWork|197|197|208|
+|SearchScreeningEvent|165|165|204|
+Canonical schemas: 1,163/1,388 (83.79%), 100% schema validity. Renamed-key stress 21/28; nullable stress 15/28. Playlist and weather exactness decrease, while restaurants, music and ratings increase. Do not describe this as uniform improvement or guaranteed nonregression. Full per-field counts, NLL and per-domain/variant metrics are in the results JSON.
+### Development and local examples
+|Panel|Previous|Main|
+|---|---:|---:|
+|Original 252 DEV complete|200/252|206/252|
+|New 48 cardinality DEV complete|15/48|32/48|
+|New two/three-value requests|0/24|14/24|
+|Earlier 8 multi-value probes|0/8|7/8|
+|54 local playground complete|42/54|44/54|
+|54 local playground schema-valid|52/54|54/54|
+Local field accuracy 124/135(91.85%); value precision 89.53%, recall 91.67%, F190.59%; missing 49/53(92.45%); unsupported values 3; all 54 stop naturally. Local precision declines from 91.03% despite higher recall/correctness. The new 48-pattern panel has only 36/48(75%) schema-valid outputs. All panels remain separate; do not pool them or claim an external leaderboard result.
+## 10. Failure audit and limitations
+All 19 newly regressed original TEST requests are schema-valid JSON. Manual primary categories: 4 changed/invented values, 3 shortened spans, 8 wrong field assignments, 2 omissions and 2 especially ambiguous source/schema contracts. Examples include Pete Murray becoming Peanut, a complete movie title losing its final phrase, a venue assigned as a movie, and an explicit city becoming null. A bare music title may not specify track versus album; broad entity_name definitions also conflict with exclusive annotation roles. Those ambiguity flags do not change benchmark golds or excuse clearer errors. Full-case evidence is retained in the report archive.
+Multiple-value coverage improves, but errors remain: Bergen and Oslo can generate Bergen, Orange and Lake. Correct JSON structure cannot establish correct extraction. The model is primarily SNIPS-domain research software; unfamiliar fields, long documents, corrections/associations, nested schemas, arbitrary-schema reliability, general chat/coding, multilingual behavior, confidence calibration and prompt-injection robustness remain unestablished. No 95% reliability or universal release-readiness claim is made.
+## 11. Packaging, verification and reproduction
+The main package retains BF16 weights, frozen tokenizer and native MLX architecture. Stock-MLX standalone verification compares all 54 raw outputs and stop reasons against the selected saved playground protocol; the receipt specifies achieved coverage/parity and correctness. This is packaging parity, not a fresh blind accuracy estimate or complete 1,444-case cross-runtime parity claim.
+Hugging Face supplies canonical model.safetensors, config, tokenizer, launcher, native_bundle.zip and model/report/results/provenance records. GitHub ships runnable source, curated JSONL records, preparation/training/evaluation code and audit evidence. Optimizer/packed-array binaries, private workspace paths, credentials and host network logs are excluded. Original workspace manifests record excluded files; provenance/SNAPSHOT_MANIFEST.json lists the actual shipped source inventory. Sanitized metadata hashes are separate from original metadata hashes.
+The additional stage has frozen paths to historical baseline receipts and prepared-data inputs. Reproduction requires the specified parent weights, original source dependencies/layout, re-packing omitted arrays and regenerating local preflight receipts; published audit receipts are not ready-to-resume optimizer state. Native inference needs none of those training dependencies. The teacher is only used during training.
+See provenance/LINEAGE.json, results/JSONLITE_RESULTS.json, pretraining/native_json_105/, examples/cases.json and reports/model-design/SFT-105_cardinality_distillation_20261014/ inside the source archive. Prior native JSON release hashes/revisions remain in provenance/RELEASE_HISTORY.json; old Git/Hub revisions remain accessible.
+## 12. Attribution and licensing
+Source code retains Apache-2.0; upstream data and dependencies retain their notices. Original SNIPS annotations are CC0; the matched cardinality examples are programmatically authored from TRAIN literals. The earlier SQuAD/Dolly/pretraining source notices remain in THIRD_PARTY_NOTICES.md. Source licensing does not automatically assign a trained-weight license; none is assigned by this update.
+Hugging Face edition: archive-relative code/evidence references resolve inside native_bundle.zip.

checkpoint_metadata.json CHANGED Viewed

@@ -51,6 +51,12 @@
         "pretraining/native_json_100/runtime.py": "9b04f8b33fceb717fa6e9290a75bcc7c8365ad6162d3f59de19e08b913a034c4",
         "pretraining/native_json_100/train.py": "63823c5bc7a4b44cb57d7cd4bead7e97ecf309b09816e4d81327a9b7d795ea72",
         "pretraining/native_json_100/verify.py": "aa1452e3994eb534f30185c69ba1fcac5dc47081a51e547c31db132e8c369118",
         "pretraining/qa_064/data.py": "a5d86a1c73285f420c728b5e8331da468eb8846086a8bbe69ce61df6e64fef7f",
         "pretraining/qa_064/evaluate.py": "b0b8bd42b4842065e3ef8ba3c1fee42c95464f0f8ec49d030e4ab4f1998f94ae",
         "pretraining/qa_064/runtime.py": "7b97df4835d03210c77da36f53340aa83d5915b9cfa7aae968aef6961422c2f1",
@@ -65,7 +71,11 @@
         "pretraining/qa_066/evaluate.py": "d998fb255daabc15b6070bffaf9adeeebdbe631e54be05b83b473f79b3c8642c",
         "pretraining/qa_066/runtime.py": "7b97df4835d03210c77da36f53340aa83d5915b9cfa7aae968aef6961422c2f1",
         "pretraining/qa_066/train.py": "3da0be266aa1f047addba8d0c696245ed71e4fde236ecfc8b383c69e8b32e783",
-        "pretraining/scripts/run_sft_100_native_json.sh": "40be750725ffaae7acb81fa03c592cea1b862dba409532ad45d9e7e64fc34fce",
         "pretraining/sft/data.py": "70348a9df1473da88330695d5409f17b929b9bc1dc79f77525272f6d87317aa1",
         "pretraining/sft/evaluate.py": "80624551512112850145ccb57d6435d7a1bae86d0b95f8f52d56b8cb6f102a76",
         "pretraining/sft/quality.py": "2abf5926a22a33b4f29c699855b24f7e459b6e631aa6ef780876ddc205f17758",
@@ -102,40 +112,44 @@
         "pretraining/sml_v1/tokenization.py": "254e1f5d60cc4498a9f1209955752db28b91839f48d3121d674f3f5ac42e52e5",
         "pretraining/sml_v1/topology_transition.py": "adc3a0a6f19e7cfb3073f4c1e0432f664e5fb709141ceb4148aca7cda08ddaab",
         "pretraining/sml_v1/train_tokenizer.py": "90c58a30a50cb625e36222825768a134aba5c81df87de8655ab95de2bd47402c",
-        "pretraining/tests/test_native_json_100.py": "00b4d10d22d7497dd14444cf6a669f9dc195f6d1b52dc7dcb679e76f7057483b",
         "reference/opensml-150m/tokenizer/bytebpe32k_v1/corpus.json": "e4c47b7b6aba08f948de7c2c7cb94c58dbd09dde2ded6ac185c60856ee686d27",
         "reference/opensml-150m/tokenizer/bytebpe32k_v1/manifest.json": "1194292c6d906ac19e6f01c6ad2b42825143fad426a22aa8856d903776701ecb",
         "reference/opensml-150m/tokenizer/bytebpe32k_v1/report.json": "e871f690eabebf82da173f0e406f38b79a10cd7d2647bb3592c3fc35ae9a2f16",
         "reference/opensml-150m/tokenizer/bytebpe32k_v1/tokenizer.json": "e095f0413a11fa91fbb58f37d00e4c6bcda59beb89b28f6f680db1436a7e8891",
-        "reference/opensml-150m/tokenizer/bytebpe32k_v1/tokenizer_config.json": "619f500553367eb76f5be00a98625130ec857b9c367561334984096c029ba2a9"
       },
-      "manifest": "f0d23b86eeba71d7f616ab630da8914aed4f5586f7dc149426b487581c833f00",
-      "parent": "[workspace]/pretraining/runs/sft_093_json_approaches_20261014/step_0000160_82ffab052730",
-      "parent_metadata": "32a380899c455449db066ef317de9dd8e80f6977bfb42a6fe03b67c5f23963f8",
-      "parent_sha256": "d552c1380c71c7c0569e9d692ec9c46c5c94bb87a7b024e56df0f0e36df4d798",
       "plan": {
         "batch_size": 32,
         "betas": [
           0.9,
           0.95
         ],
         "epochs": 1,
         "eps": 1e-08,
         "eval_steps": [
           0,
-          100,
-          200,
-          300,
-          469
         ],
-        "final_lr": 1e-06,
         "grad_clip": 1.0,
         "max_tokens": 256,
         "microbatch": 2,
         "padding_policy": "micro_multiple128",
-        "peak_lr": 5e-06,
-        "seed": 20261014,
-        "warmup_updates": 20,
         "weight_decay": 0.01
       },
       "versions": {
@@ -145,10 +159,10 @@
         "tokenizers": "0.22.2"
       }
     },
-    "order_hash": "76518047ebfc78beb886d78322635ee275acdb6ad479a5d722bb871be3c89961",
-    "records": 15000,
     "serialization": "native-sft-chat-v1",
-    "total_steps": 469
   },
   "recipe": {
     "model": {
@@ -169,6 +183,6 @@
       "vocab_size": 32000
     }
   },
-  "step": 469,
-  "tokens": 4634126
 }

         "pretraining/native_json_100/runtime.py": "9b04f8b33fceb717fa6e9290a75bcc7c8365ad6162d3f59de19e08b913a034c4",
         "pretraining/native_json_100/train.py": "63823c5bc7a4b44cb57d7cd4bead7e97ecf309b09816e4d81327a9b7d795ea72",
         "pretraining/native_json_100/verify.py": "aa1452e3994eb534f30185c69ba1fcac5dc47081a51e547c31db132e8c369118",
+        "pretraining/native_json_105/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+        "pretraining/native_json_105/data.py": "b4934d571851f934164a2f0dce310486dc70e45c8222f8ae5c91a9fd987be303",
+        "pretraining/native_json_105/evaluate.py": "bf439186ebf0819f54e2e1dafb120e3abc0e05b53e2367f3adb6abc09f66817b",
+        "pretraining/native_json_105/runtime.py": "cd00debd869794c50af2856e51437199bbe2221bfb1b7192e36332167ef1ee24",
+        "pretraining/native_json_105/train.py": "fec330fbcf8d6f3b76cc4c86a30bcf00f7215cbc57b1af31ebcc05c69f5c3d4d",
+        "pretraining/native_json_105/verify.py": "e21a616dba7b402aeaea40c204678fdbfc16dda97606d74179c7a213cfb57d78",
         "pretraining/qa_064/data.py": "a5d86a1c73285f420c728b5e8331da468eb8846086a8bbe69ce61df6e64fef7f",
         "pretraining/qa_064/evaluate.py": "b0b8bd42b4842065e3ef8ba3c1fee42c95464f0f8ec49d030e4ab4f1998f94ae",
         "pretraining/qa_064/runtime.py": "7b97df4835d03210c77da36f53340aa83d5915b9cfa7aae968aef6961422c2f1",
         "pretraining/qa_066/evaluate.py": "d998fb255daabc15b6070bffaf9adeeebdbe631e54be05b83b473f79b3c8642c",
         "pretraining/qa_066/runtime.py": "7b97df4835d03210c77da36f53340aa83d5915b9cfa7aae968aef6961422c2f1",
         "pretraining/qa_066/train.py": "3da0be266aa1f047addba8d0c696245ed71e4fde236ecfc8b383c69e8b32e783",
+        "pretraining/runs/sft_100_native_json_20261014/evaluations/development_0000469.identity.json": "77385825353ef560758a91527a4d47113b319694097fb62147d3e88cb123279e",
+        "pretraining/runs/sft_100_native_json_20261014/evaluations/development_0000469.json": "bdb12afb6ab0a6d0066894f3d18c8fe6e81aeaa4c4e7cfe579452d4273aa0585",
+        "pretraining/runs/sft_100_native_json_20261014/final-test-identity.json": "80ac5bd9eb25dcdd99cb1bf2f4cbaa20e0c79ef334b4c37e310af44a2cb2cb48",
+        "pretraining/runs/sft_100_native_json_20261014/final-test.json": "8f0efd027a57d34f8e22f2ac5d2d54b72a1103a2d853b3704514947559c2e3b0",
+        "pretraining/scripts/run_sft_105_cardinality_distillation.sh": "4d4ca56869958045f7504b08684d1610422cee323e146648af01c75a908b60eb",
         "pretraining/sft/data.py": "70348a9df1473da88330695d5409f17b929b9bc1dc79f77525272f6d87317aa1",
         "pretraining/sft/evaluate.py": "80624551512112850145ccb57d6435d7a1bae86d0b95f8f52d56b8cb6f102a76",
         "pretraining/sft/quality.py": "2abf5926a22a33b4f29c699855b24f7e459b6e631aa6ef780876ddc205f17758",
         "pretraining/sml_v1/tokenization.py": "254e1f5d60cc4498a9f1209955752db28b91839f48d3121d674f3f5ac42e52e5",
         "pretraining/sml_v1/topology_transition.py": "adc3a0a6f19e7cfb3073f4c1e0432f664e5fb709141ceb4148aca7cda08ddaab",
         "pretraining/sml_v1/train_tokenizer.py": "90c58a30a50cb625e36222825768a134aba5c81df87de8655ab95de2bd47402c",
+        "pretraining/tests/test_native_json_105.py": "698d9410a900dee6b81dede19f20ea06b721692b2d5a09ebb4ee646d9786701a",
         "reference/opensml-150m/tokenizer/bytebpe32k_v1/corpus.json": "e4c47b7b6aba08f948de7c2c7cb94c58dbd09dde2ded6ac185c60856ee686d27",
         "reference/opensml-150m/tokenizer/bytebpe32k_v1/manifest.json": "1194292c6d906ac19e6f01c6ad2b42825143fad426a22aa8856d903776701ecb",
         "reference/opensml-150m/tokenizer/bytebpe32k_v1/report.json": "e871f690eabebf82da173f0e406f38b79a10cd7d2647bb3592c3fc35ae9a2f16",
         "reference/opensml-150m/tokenizer/bytebpe32k_v1/tokenizer.json": "e095f0413a11fa91fbb58f37d00e4c6bcda59beb89b28f6f680db1436a7e8891",
+        "reference/opensml-150m/tokenizer/bytebpe32k_v1/tokenizer_config.json": "619f500553367eb76f5be00a98625130ec857b9c367561334984096c029ba2a9",
+        "reports/model-design/SFT-105_cardinality_distillation_20261014/parent-development.json": "581a70411556d0af2e9aa6868af29c614d89f94012179c6b14b897b0afa0dfe0"
       },
+      "manifest": "3680d620c3cefd00473ee5fcabe60b5e3a1351ef5b2d8c4ef763efec0081a218",
+      "parent": "[workspace]/pretraining/runs/sft_100_native_json_20261014/step_0000469_25528943984d",
+      "parent_metadata": "3864a0eaa43015d286e4da9327b0d3cea4fa4f5b80f336baabe2a549379e2e36",
+      "parent_sha256": "656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4",
       "plan": {
         "batch_size": 32,
         "betas": [
           0.9,
           0.95
         ],
+        "distillation_temperature": 1.0,
         "epochs": 1,
         "eps": 1e-08,
         "eval_steps": [
           0,
+          32,
+          64,
+          96
         ],
+        "final_lr": 2e-07,
         "grad_clip": 1.0,
         "max_tokens": 256,
         "microbatch": 2,
+        "order_policy": "prepared_prefix",
         "padding_policy": "micro_multiple128",
+        "peak_lr": 2e-06,
+        "replay_kl_weight": 1.0,
+        "seed": 20261119,
+        "total_updates": 96,
+        "warmup_updates": 6,
         "weight_decay": 0.01
       },
       "versions": {
         "tokenizers": "0.22.2"
       }
     },
+    "order_hash": "8e0ef3d752534f0aa8e45465436974b77384cc65b3facba7f32c8bc29bcf0750",
+    "records": 2048,
     "serialization": "native-sft-chat-v1",
+    "total_steps": 96
   },
   "recipe": {
     "model": {
       "vocab_size": 32000
     }
   },
+  "step": 64,
+  "tokens": 600558
 }

checkpoint_provenance.json CHANGED Viewed

@@ -46,13 +46,25 @@
       "weight_bytes": 351242342,
       "original_metadata_sha256": "3864a0eaa43015d286e4da9327b0d3cea4fa4f5b80f336baabe2a549379e2e36",
       "published_metadata_sha256": "5ecf9a8850aef6cc75add6bacda880648e72de1b38089d585e237bbce828c2c3"
     }
   ],
   "weights_in_git": false,
-  "public_weight_location": null,
   "model_weight_license": "not yet assigned",
-  "scope": "Stage A/132005 -> SFT-065/469 -> SFT-068/379 -> SFT-093/160 -> SFT-100/469. Only those stages are inherited. Final native JSON targets are directly supervised.",
-  "selected_stage": "sft-100",
-  "selected_update": 469,
-  "weight_location": "https://huggingface.co/OpenSML/OpenSML-JSONLite-175M"
 }

       "weight_bytes": 351242342,
       "original_metadata_sha256": "3864a0eaa43015d286e4da9327b0d3cea4fa4f5b80f336baabe2a549379e2e36",
       "published_metadata_sha256": "5ecf9a8850aef6cc75add6bacda880648e72de1b38089d585e237bbce828c2c3"
+    },
+    {
+      "stage": "sft-105",
+      "public_stage": "additional extraction SFT",
+      "step": 64,
+      "tokens": 600558,
+      "records": 2048,
+      "weight_sha256": "655fb5cfba009163cd8a7f7fe7b170b998bb452c217b1b473cfe10b33de0ccd1",
+      "weight_bytes": 351242342,
+      "original_metadata_sha256": "8b8c0049d731f8124eaec78a213d05c24a520bc4f190dc82975d763f43351bbc",
+      "published_metadata_sha256": "49d177fb4606673f789fbf06c694643c44504672a9f0c3f75f7eae776cb4590c"
     }
   ],
   "weights_in_git": false,
+  "public_weight_location": "https://huggingface.co/OpenSML/OpenSML-JSONLite-175M",
   "model_weight_license": "not yet assigned",
+  "scope": "Stage A -> grounded QA -> requested-field extraction -> focused extraction -> native JSON foundation -> cardinality/replay distillation continuation. Only the selected prefix of the final continuation is inherited; complete historical update counts are in stages.",
+  "selected_stage": "additional extraction SFT",
+  "selected_update": 64,
+  "weight_location": "https://huggingface.co/OpenSML/OpenSML-JSONLite-175M",
+  "public_checkpoint_name": "main"
 }

complete-request-exact.svg CHANGED Viewed

config.json CHANGED Viewed

@@ -18,7 +18,7 @@
     "rope_base": 10000.0,
     "vocab_size": 32000
   },
-  "weight_sha256": "656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4",
   "serialization": "native-sft-chat-v1",
   "output_mode": "native-json",
   "context_limit": 2048,
@@ -29,6 +29,8 @@
   "json_assembly": false,
   "output_repair": false,
   "constrained_decoding": false,
-  "selected_stage": "sft-100",
-  "selected_update": 469
 }

     "rope_base": 10000.0,
     "vocab_size": 32000
   },
+  "weight_sha256": "655fb5cfba009163cd8a7f7fe7b170b998bb452c217b1b473cfe10b33de0ccd1",
   "serialization": "native-sft-chat-v1",
   "output_mode": "native-json",
   "context_limit": 2048,
   "json_assembly": false,
   "output_repair": false,
   "constrained_decoding": false,
+  "selected_stage": "additional extraction SFT",
+  "selected_update": 64,
+  "checkpoint_name": "main",
+  "inference_teacher_required": false
 }

inference.py CHANGED Viewed

@@ -12,7 +12,7 @@ import tempfile
 import zipfile
 ROOT = Path(__file__).resolve().parent
-BUNDLE_SHA256 = "4ea6280d2e60da34e59757a4d6ab95aaf7811050c68c047cf5b98d2f001e9b78"
 def main():
     archive = ROOT / "native_bundle.zip"

 import zipfile
 ROOT = Path(__file__).resolve().parent
+BUNDLE_SHA256 = "a1b63ed6286d6c4f8bc4f7526f4a11fc46f936830daaac22224bfa0c4892755b"
 def main():
     archive = ROOT / "native_bundle.zip"

model.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4
 size 351242342

 version https://git-lfs.github.com/spec/v1
+oid sha256:655fb5cfba009163cd8a7f7fe7b170b998bb452c217b1b473cfe10b33de0ccd1
 size 351242342

native_bundle.zip CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4ea6280d2e60da34e59757a4d6ab95aaf7811050c68c047cf5b98d2f001e9b78
-size 5607781

 version https://git-lfs.github.com/spec/v1
+oid sha256:a1b63ed6286d6c4f8bc4f7526f4a11fc46f936830daaac22224bfa0c4892755b
+size 7638568