Upload complete native JSON model, runtime and verified examples

Native source-text + JSON Schema generation weights, tokenizer, MLX runtime, training report, evaluation evidence, checksums and four verified playground examples.

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+pretraining-training-loss.png filter=lfs diff=lfs merge=lfs -text

CITATION.cff

@@ -0,0 +1,12 @@
+cff-version: 1.2.0
+message: "Please cite this source repository and the selected checkpoint and native inference identity when reporting OpenSML-JSONLite-175M."
+title: "OpenSML-JSONLite-175M: Native JSON Generation from Source Text and Schema"
+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"
+license: Apache-2.0
+abstract: "Source, tokenizer, training provenance and observed evaluation records for a 175.6M-parameter native text-to-JSON model. Source license only; weights are not included."

INFERENCE_VERIFICATION.json

@@ -0,0 +1,454 @@
+{
+  "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"
+}

JSONLITE_RESULTS.json

@@ -0,0 +1,1162 @@
+{
+  "model": "OpenSML-JSONLite-175M-Instruct",
+  "selected_stage": "sft-100",
+  "step": 469,
+  "model_sha256": "656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4",
+  "output_mode": "native-json",
+  "development": {
+    "metrics": {
+      "all": {
+        "absent_accuracy": 0.9294117647058824,
+        "count": 252,
+        "counts": {
+          "absent_correct": 237,
+          "absent_total": 255,
+          "exact": 200,
+          "false_abstention": 25,
+          "field_correct": 546,
+          "fields": 604,
+          "fn": 40,
+          "fp": 29,
+          "json_parse_ok": 251,
+          "missed_abstention": 18,
+          "natural_stop": 252,
+          "nonrepetitive_nonempty": 252,
+          "predicted_values": 339,
+          "present_correct": 309,
+          "present_total": 349,
+          "schema_ok": 251,
+          "token_limited": 0,
+          "tp": 310,
+          "unsupported_values": 4
+        },
+        "exact": 0.7936507936507936,
+        "f1": 0.899854862119013,
+        "false_abstention_rate": 0.07163323782234957,
+        "field_accuracy": 0.9039735099337748,
+        "json_parse_rate": 0.996031746031746,
+        "missed_abstention_rate": 0.07058823529411765,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.03873514435736791,
+        "precision": 0.9144542772861357,
+        "present_accuracy": 0.8853868194842407,
+        "recall": 0.8857142857142857,
+        "schema_valid_rate": 0.996031746031746,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.011799410029498525
+      },
+      "domain/AddToPlaylist": {
+        "absent_accuracy": 0.9024390243902439,
+        "count": 36,
+        "counts": {
+          "absent_correct": 37,
+          "absent_total": 41,
+          "exact": 20,
+          "false_abstention": 9,
+          "field_correct": 72,
+          "fields": 91,
+          "fn": 15,
+          "fp": 7,
+          "json_parse_ok": 35,
+          "missed_abstention": 4,
+          "natural_stop": 36,
+          "nonrepetitive_nonempty": 36,
+          "predicted_values": 42,
+          "present_correct": 35,
+          "present_total": 50,
+          "schema_ok": 35,
+          "token_limited": 0,
+          "tp": 35,
+          "unsupported_values": 2
+        },
+        "exact": 0.5555555555555556,
+        "f1": 0.7608695652173915,
+        "false_abstention_rate": 0.18,
+        "field_accuracy": 0.7912087912087912,
+        "json_parse_rate": 0.9722222222222222,
+        "missed_abstention_rate": 0.0975609756097561,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.08232390080795288,
+        "precision": 0.8333333333333334,
+        "present_accuracy": 0.7,
+        "recall": 0.7,
+        "schema_valid_rate": 0.9722222222222222,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.047619047619047616
+      },
+      "domain/BookRestaurant": {
+        "absent_accuracy": 0.8604651162790697,
+        "count": 36,
+        "counts": {
+          "absent_correct": 37,
+          "absent_total": 43,
+          "exact": 24,
+          "false_abstention": 4,
+          "field_correct": 82,
+          "fields": 95,
+          "fn": 7,
+          "fp": 9,
+          "json_parse_ok": 36,
+          "missed_abstention": 6,
+          "natural_stop": 36,
+          "nonrepetitive_nonempty": 36,
+          "predicted_values": 54,
+          "present_correct": 45,
+          "present_total": 52,
+          "schema_ok": 36,
+          "token_limited": 0,
+          "tp": 45,
+          "unsupported_values": 2
+        },
+        "exact": 0.6666666666666666,
+        "f1": 0.8490566037735849,
+        "false_abstention_rate": 0.07692307692307693,
+        "field_accuracy": 0.8631578947368421,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.13953488372093023,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.0473247706718717,
+        "precision": 0.8333333333333334,
+        "present_accuracy": 0.8653846153846154,
+        "recall": 0.8653846153846154,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.037037037037037035
+      },
+      "domain/GetWeather": {
+        "absent_accuracy": 0.925,
+        "count": 36,
+        "counts": {
+          "absent_correct": 37,
+          "absent_total": 40,
+          "exact": 28,
+          "false_abstention": 5,
+          "field_correct": 85,
+          "fields": 94,
+          "fn": 6,
+          "fp": 3,
+          "json_parse_ok": 36,
+          "missed_abstention": 3,
+          "natural_stop": 36,
+          "nonrepetitive_nonempty": 36,
+          "predicted_values": 52,
+          "present_correct": 48,
+          "present_total": 54,
+          "schema_ok": 36,
+          "token_limited": 0,
+          "tp": 49,
+          "unsupported_values": 0
+        },
+        "exact": 0.7777777777777778,
+        "f1": 0.9158878504672897,
+        "false_abstention_rate": 0.09259259259259259,
+        "field_accuracy": 0.9042553191489362,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.075,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.044021505151047476,
+        "precision": 0.9423076923076923,
+        "present_accuracy": 0.8888888888888888,
+        "recall": 0.8909090909090909,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.0
+      },
+      "domain/PlayMusic": {
+        "absent_accuracy": 0.9767441860465116,
+        "count": 36,
+        "counts": {
+          "absent_correct": 42,
+          "absent_total": 43,
+          "exact": 30,
+          "false_abstention": 6,
+          "field_correct": 81,
+          "fields": 88,
+          "fn": 6,
+          "fp": 1,
+          "json_parse_ok": 36,
+          "missed_abstention": 1,
+          "natural_stop": 36,
+          "nonrepetitive_nonempty": 36,
+          "predicted_values": 40,
+          "present_correct": 39,
+          "present_total": 45,
+          "schema_ok": 36,
+          "token_limited": 0,
+          "tp": 39,
+          "unsupported_values": 0
+        },
+        "exact": 0.8333333333333334,
+        "f1": 0.9176470588235294,
+        "false_abstention_rate": 0.13333333333333333,
+        "field_accuracy": 0.9204545454545454,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.023255813953488372,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.04131003013268734,
+        "precision": 0.975,
+        "present_accuracy": 0.8666666666666667,
+        "recall": 0.8666666666666667,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.0
+      },
+      "domain/RateBook": {
+        "absent_accuracy": 0.9767441860465116,
+        "count": 36,
+        "counts": {
+          "absent_correct": 42,
+          "absent_total": 43,
+          "exact": 34,
+          "false_abstention": 0,
+          "field_correct": 94,
+          "fields": 96,
+          "fn": 1,
+          "fp": 2,
+          "json_parse_ok": 36,
+          "missed_abstention": 1,
+          "natural_stop": 36,
+          "nonrepetitive_nonempty": 36,
+          "predicted_values": 54,
+          "present_correct": 52,
+          "present_total": 53,
+          "schema_ok": 36,
+          "token_limited": 0,
+          "tp": 52,
+          "unsupported_values": 0
+        },
+        "exact": 0.9444444444444444,
+        "f1": 0.9719626168224299,
+        "false_abstention_rate": 0.0,
+        "field_accuracy": 0.9791666666666666,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.023255813953488372,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.01588757585947557,
+        "precision": 0.9629629629629629,
+        "present_accuracy": 0.9811320754716981,
+        "recall": 0.9811320754716981,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.0
+      },
+      "domain/SearchCreativeWork": {
+        "absent_accuracy": 1.0,
+        "count": 36,
+        "counts": {
+          "absent_correct": 4,
+          "absent_total": 4,
+          "exact": 33,
+          "false_abstention": 1,
+          "field_correct": 44,
+          "fields": 47,
+          "fn": 3,
+          "fp": 2,
+          "json_parse_ok": 36,
+          "missed_abstention": 0,
+          "natural_stop": 36,
+          "nonrepetitive_nonempty": 36,
+          "predicted_values": 42,
+          "present_correct": 40,
+          "present_total": 43,
+          "schema_ok": 36,
+          "token_limited": 0,
+          "tp": 40,
+          "unsupported_values": 0
+        },
+        "exact": 0.9166666666666666,
+        "f1": 0.9411764705882352,
+        "false_abstention_rate": 0.023255813953488372,
+        "field_accuracy": 0.9361702127659575,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.0,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.017151051737730287,
+        "precision": 0.9523809523809523,
+        "present_accuracy": 0.9302325581395349,
+        "recall": 0.9302325581395349,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.0
+      },
+      "domain/SearchScreeningEvent": {
+        "absent_accuracy": 0.926829268292683,
+        "count": 36,
+        "counts": {
+          "absent_correct": 38,
+          "absent_total": 41,
+          "exact": 31,
+          "false_abstention": 0,
+          "field_correct": 88,
+          "fields": 93,
+          "fn": 2,
+          "fp": 5,
+          "json_parse_ok": 36,
+          "missed_abstention": 3,
+          "natural_stop": 36,
+          "nonrepetitive_nonempty": 36,
+          "predicted_values": 55,
+          "present_correct": 50,
+          "present_total": 52,
+          "schema_ok": 36,
+          "token_limited": 0,
+          "tp": 50,
+          "unsupported_values": 0
+        },
+        "exact": 0.8611111111111112,
+        "f1": 0.9345794392523366,
+        "false_abstention_rate": 0.0,
+        "field_accuracy": 0.946236559139785,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.07317073170731707,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.023127176140810154,
+        "precision": 0.9090909090909091,
+        "present_accuracy": 0.9615384615384616,
+        "recall": 0.9615384615384616,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.0
+      },
+      "macro_exact": 0.7936507936507936,
+      "paired_requests": {
+        "accuracy": 0.7040816326530612,
+        "both_correct": 69,
+        "count": 98
+      },
+      "variant/alias": {
+        "absent_accuracy": 0.96875,
+        "count": 28,
+        "counts": {
+          "absent_correct": 31,
+          "absent_total": 32,
+          "exact": 24,
+          "false_abstention": 3,
+          "field_correct": 64,
+          "fields": 68,
+          "fn": 3,
+          "fp": 1,
+          "json_parse_ok": 28,
+          "missed_abstention": 1,
+          "natural_stop": 28,
+          "nonrepetitive_nonempty": 28,
+          "predicted_values": 34,
+          "present_correct": 33,
+          "present_total": 36,
+          "schema_ok": 28,
+          "token_limited": 0,
+          "tp": 33,
+          "unsupported_values": 0
+        },
+        "exact": 0.8571428571428571,
+        "f1": 0.9428571428571428,
+        "false_abstention_rate": 0.08333333333333333,
+        "field_accuracy": 0.9411764705882353,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.03125,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.058719543459509235,
+        "precision": 0.9705882352941176,
+        "present_accuracy": 0.9166666666666666,
+        "recall": 0.9166666666666666,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.0
+      },
+      "variant/canonical": {
+        "absent_accuracy": 0.9319371727748691,
+        "count": 196,
+        "counts": {
+          "absent_correct": 178,
+          "absent_total": 191,
+          "exact": 157,
+          "false_abstention": 17,
+          "field_correct": 426,
+          "fields": 468,
+          "fn": 29,
+          "fp": 24,
+          "json_parse_ok": 196,
+          "missed_abstention": 13,
+          "natural_stop": 196,
+          "nonrepetitive_nonempty": 196,
+          "predicted_values": 273,
+          "present_correct": 248,
+          "present_total": 277,
+          "schema_ok": 196,
+          "token_limited": 0,
+          "tp": 249,
+          "unsupported_values": 3
+        },
+        "exact": 0.8010204081632653,
+        "f1": 0.9038112522686025,
+        "false_abstention_rate": 0.061371841155234655,
+        "field_accuracy": 0.9102564102564102,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.06806282722513089,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.02958408025570323,
+        "precision": 0.9120879120879121,
+        "present_accuracy": 0.8953068592057761,
+        "recall": 0.89568345323741,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.01098901098901099
+      },
+      "variant/nullable": {
+        "absent_accuracy": 0.875,
+        "count": 28,
+        "counts": {
+          "absent_correct": 28,
+          "absent_total": 32,
+          "exact": 19,
+          "false_abstention": 5,
+          "field_correct": 56,
+          "fields": 68,
+          "fn": 8,
+          "fp": 4,
+          "json_parse_ok": 27,
+          "missed_abstention": 4,
+          "natural_stop": 28,
+          "nonrepetitive_nonempty": 28,
+          "predicted_values": 32,
+          "present_correct": 28,
+          "present_total": 36,
+          "schema_ok": 27,
+          "token_limited": 0,
+          "tp": 28,
+          "unsupported_values": 1
+        },
+        "exact": 0.6785714285714286,
+        "f1": 0.823529411764706,
+        "false_abstention_rate": 0.1388888888888889,
+        "field_accuracy": 0.8235294117647058,
+        "json_parse_rate": 0.9642857142857143,
+        "missed_abstention_rate": 0.125,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.08280819396687937,
+        "precision": 0.875,
+        "present_accuracy": 0.7777777777777778,
+        "recall": 0.7777777777777778,
+        "schema_valid_rate": 0.9642857142857143,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.03125
+      }
+    },
+    "protocol": {
+      "constrained_decoding": false,
+      "json_assembly": false,
+      "max_tokens": 256,
+      "model_calls_per_request": 1,
+      "output_repair": false,
+      "prompt_truncation": false,
+      "repetition_penalty": 1,
+      "scorer": "strict-raw-json100-v1",
+      "serialization": "native-sft-chat-v1",
+      "temperature": 0
+    },
+    "scope": "Raw model-generated JSON, source-span annotation agreement. Duplicate keys, nonfinite JSON, fences, trailing prose and schema violations fail exact correctness. Flat string arrays/nullable strings in seven SNIPS domains; observed research panels, not a new blind or external benchmark."
+  },
+  "test": {
+    "metrics": {
+      "all": {
+        "absent_accuracy": 0.9246395806028833,
+        "count": 1444,
+        "counts": {
+          "absent_correct": 1411,
+          "absent_total": 1526,
+          "exact": 1191,
+          "false_abstention": 111,
+          "field_correct": 3193,
+          "fields": 3497,
+          "fn": 189,
+          "fp": 186,
+          "json_parse_ok": 1442,
+          "missed_abstention": 115,
+          "natural_stop": 1444,
+          "nonrepetitive_nonempty": 1444,
+          "predicted_values": 1968,
+          "present_correct": 1782,
+          "present_total": 1971,
+          "schema_ok": 1441,
+          "token_limited": 0,
+          "tp": 1782,
+          "unsupported_values": 30
+        },
+        "exact": 0.824792243767313,
+        "f1": 0.9047981721249048,
+        "false_abstention_rate": 0.0563165905631659,
+        "field_accuracy": 0.9130683442951101,
+        "json_parse_rate": 0.9986149584487535,
+        "missed_abstention_rate": 0.07536041939711664,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.028334847962383555,
+        "precision": 0.9054878048780488,
+        "present_accuracy": 0.9041095890410958,
+        "recall": 0.9041095890410958,
+        "schema_valid_rate": 0.9979224376731302,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.01524390243902439
+      },
+      "domain/AddToPlaylist": {
+        "absent_accuracy": 0.9110169491525424,
+        "count": 206,
+        "counts": {
+          "absent_correct": 215,
+          "absent_total": 236,
+          "exact": 168,
+          "false_abstention": 13,
+          "field_correct": 475,
+          "fields": 526,
+          "fn": 30,
+          "fp": 38,
+          "json_parse_ok": 206,
+          "missed_abstention": 21,
+          "natural_stop": 206,
+          "nonrepetitive_nonempty": 206,
+          "predicted_values": 298,
+          "present_correct": 260,
+          "present_total": 290,
+          "schema_ok": 206,
+          "token_limited": 0,
+          "tp": 260,
+          "unsupported_values": 13
+        },
+        "exact": 0.8155339805825242,
+        "f1": 0.8843537414965987,
+        "false_abstention_rate": 0.04482758620689655,
+        "field_accuracy": 0.903041825095057,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.08898305084745763,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.04326067907575776,
+        "precision": 0.87248322147651,
+        "present_accuracy": 0.896551724137931,
+        "recall": 0.896551724137931,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.0436241610738255
+      },
+      "domain/BookRestaurant": {
+        "absent_accuracy": 0.908,
+        "count": 208,
+        "counts": {
+          "absent_correct": 227,
+          "absent_total": 250,
+          "exact": 156,
+          "false_abstention": 21,
+          "field_correct": 506,
+          "fields": 569,
+          "fn": 40,
+          "fp": 42,
+          "json_parse_ok": 208,
+          "missed_abstention": 23,
+          "natural_stop": 208,
+          "nonrepetitive_nonempty": 208,
+          "predicted_values": 321,
+          "present_correct": 279,
+          "present_total": 319,
+          "schema_ok": 208,
+          "token_limited": 0,
+          "tp": 279,
+          "unsupported_values": 4
+        },
+        "exact": 0.75,
+        "f1": 0.8718750000000001,
+        "false_abstention_rate": 0.06583072100313479,
+        "field_accuracy": 0.8892794376098418,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.092,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.032639438736133834,
+        "precision": 0.8691588785046729,
+        "present_accuracy": 0.8746081504702194,
+        "recall": 0.8746081504702194,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.012461059190031152
+      },
+      "domain/GetWeather": {
+        "absent_accuracy": 0.9349593495934959,
+        "count": 206,
+        "counts": {
+          "absent_correct": 230,
+          "absent_total": 246,
+          "exact": 168,
+          "false_abstention": 15,
+          "field_correct": 479,
+          "fields": 525,
+          "fn": 30,
+          "fp": 27,
+          "json_parse_ok": 205,
+          "missed_abstention": 16,
+          "natural_stop": 206,
+          "nonrepetitive_nonempty": 206,
+          "predicted_values": 276,
+          "present_correct": 249,
+          "present_total": 279,
+          "schema_ok": 204,
+          "token_limited": 0,
+          "tp": 249,
+          "unsupported_values": 2
+        },
+        "exact": 0.8155339805825242,
+        "f1": 0.8972972972972973,
+        "false_abstention_rate": 0.053763440860215055,
+        "field_accuracy": 0.9123809523809524,
+        "json_parse_rate": 0.9951456310679612,
+        "missed_abstention_rate": 0.06504065040650407,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.028848352069651157,
+        "precision": 0.9021739130434783,
+        "present_accuracy": 0.8924731182795699,
+        "recall": 0.8924731182795699,
+        "schema_valid_rate": 0.9902912621359223,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.007246376811594203
+      },
+      "domain/PlayMusic": {
+        "absent_accuracy": 0.94140625,
+        "count": 208,
+        "counts": {
+          "absent_correct": 241,
+          "absent_total": 256,
+          "exact": 149,
+          "false_abstention": 44,
+          "field_correct": 451,
+          "fields": 523,
+          "fn": 57,
+          "fp": 28,
+          "json_parse_ok": 208,
+          "missed_abstention": 15,
+          "natural_stop": 208,
+          "nonrepetitive_nonempty": 208,
+          "predicted_values": 238,
+          "present_correct": 210,
+          "present_total": 267,
+          "schema_ok": 208,
+          "token_limited": 0,
+          "tp": 210,
+          "unsupported_values": 2
+        },
+        "exact": 0.7163461538461539,
+        "f1": 0.8316831683168316,
+        "false_abstention_rate": 0.1647940074906367,
+        "field_accuracy": 0.8623326959847036,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.05859375,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.04051304714644525,
+        "precision": 0.8823529411764706,
+        "present_accuracy": 0.7865168539325843,
+        "recall": 0.7865168539325843,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.008403361344537815
+      },
+      "domain/RateBook": {
+        "absent_accuracy": 0.9552845528455285,
+        "count": 204,
+        "counts": {
+          "absent_correct": 235,
+          "absent_total": 246,
+          "exact": 188,
+          "false_abstention": 2,
+          "field_correct": 548,
+          "fields": 567,
+          "fn": 8,
+          "fp": 14,
+          "json_parse_ok": 203,
+          "missed_abstention": 11,
+          "natural_stop": 204,
+          "nonrepetitive_nonempty": 204,
+          "predicted_values": 327,
+          "present_correct": 313,
+          "present_total": 321,
+          "schema_ok": 203,
+          "token_limited": 0,
+          "tp": 313,
+          "unsupported_values": 2
+        },
+        "exact": 0.9215686274509803,
+        "f1": 0.9660493827160493,
+        "false_abstention_rate": 0.006230529595015576,
+        "field_accuracy": 0.9664902998236331,
+        "json_parse_rate": 0.9950980392156863,
+        "missed_abstention_rate": 0.044715447154471545,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.011863740803485545,
+        "precision": 0.9571865443425076,
+        "present_accuracy": 0.9750778816199377,
+        "recall": 0.9750778816199377,
+        "schema_valid_rate": 0.9950980392156863,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.0061162079510703364
+      },
+      "domain/SearchCreativeWork": {
+        "absent_accuracy": 0.9230769230769231,
+        "count": 208,
+        "counts": {
+          "absent_correct": 36,
+          "absent_total": 39,
+          "exact": 197,
+          "false_abstention": 2,
+          "field_correct": 275,
+          "fields": 286,
+          "fn": 8,
+          "fp": 9,
+          "json_parse_ok": 208,
+          "missed_abstention": 3,
+          "natural_stop": 208,
+          "nonrepetitive_nonempty": 208,
+          "predicted_values": 248,
+          "present_correct": 239,
+          "present_total": 247,
+          "schema_ok": 208,
+          "token_limited": 0,
+          "tp": 239,
+          "unsupported_values": 0
+        },
+        "exact": 0.9471153846153846,
+        "f1": 0.9656565656565657,
+        "false_abstention_rate": 0.008097165991902834,
+        "field_accuracy": 0.9615384615384616,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.07692307692307693,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.019352024208483972,
+        "precision": 0.9637096774193549,
+        "present_accuracy": 0.9676113360323887,
+        "recall": 0.9676113360323887,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.0
+      },
+      "domain/SearchScreeningEvent": {
+        "absent_accuracy": 0.8972332015810277,
+        "count": 204,
+        "counts": {
+          "absent_correct": 227,
+          "absent_total": 253,
+          "exact": 165,
+          "false_abstention": 14,
+          "field_correct": 459,
+          "fields": 501,
+          "fn": 16,
+          "fp": 28,
+          "json_parse_ok": 204,
+          "missed_abstention": 26,
+          "natural_stop": 204,
+          "nonrepetitive_nonempty": 204,
+          "predicted_values": 260,
+          "present_correct": 232,
+          "present_total": 248,
+          "schema_ok": 204,
+          "token_limited": 0,
+          "tp": 232,
+          "unsupported_values": 7
+        },
+        "exact": 0.8088235294117647,
+        "f1": 0.9133858267716536,
+        "false_abstention_rate": 0.056451612903225805,
+        "field_accuracy": 0.9161676646706587,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.10276679841897234,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.021568229504291415,
+        "precision": 0.8923076923076924,
+        "present_accuracy": 0.9354838709677419,
+        "recall": 0.9354838709677419,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.026923076923076925
+      },
+      "macro_exact": 0.8249888080699045,
+      "paired_requests": {
+        "accuracy": 0.7219020172910663,
+        "both_correct": 501,
+        "count": 694
+      },
+      "variant/alias": {
+        "absent_accuracy": 0.8275862068965517,
+        "count": 28,
+        "counts": {
+          "absent_correct": 24,
+          "absent_total": 29,
+          "exact": 19,
+          "false_abstention": 2,
+          "field_correct": 62,
+          "fields": 75,
+          "fn": 8,
+          "fp": 7,
+          "json_parse_ok": 27,
+          "missed_abstention": 5,
+          "natural_stop": 28,
+          "nonrepetitive_nonempty": 28,
+          "predicted_values": 45,
+          "present_correct": 38,
+          "present_total": 46,
+          "schema_ok": 26,
+          "token_limited": 0,
+          "tp": 38,
+          "unsupported_values": 2
+        },
+        "exact": 0.6785714285714286,
+        "f1": 0.8351648351648352,
+        "false_abstention_rate": 0.043478260869565216,
+        "field_accuracy": 0.8266666666666667,
+        "json_parse_rate": 0.9642857142857143,
+        "missed_abstention_rate": 0.1724137931034483,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.05925792032836138,
+        "precision": 0.8444444444444444,
+        "present_accuracy": 0.8260869565217391,
+        "recall": 0.8260869565217391,
+        "schema_valid_rate": 0.9285714285714286,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.044444444444444446
+      },
+      "variant/canonical": {
+        "absent_accuracy": 0.9305177111716622,
+        "count": 1388,
+        "counts": {
+          "absent_correct": 1366,
+          "absent_total": 1468,
+          "exact": 1156,
+          "false_abstention": 103,
+          "field_correct": 3074,
+          "fields": 3347,
+          "fn": 171,
+          "fp": 170,
+          "json_parse_ok": 1388,
+          "missed_abstention": 102,
+          "natural_stop": 1388,
+          "nonrepetitive_nonempty": 1388,
+          "predicted_values": 1878,
+          "present_correct": 1708,
+          "present_total": 1879,
+          "schema_ok": 1388,
+          "token_limited": 0,
+          "tp": 1708,
+          "unsupported_values": 24
+        },
+        "exact": 0.8328530259365994,
+        "f1": 0.909236092627096,
+        "false_abstention_rate": 0.05481639169771155,
+        "field_accuracy": 0.9184344188825814,
+        "json_parse_rate": 1.0,
+        "missed_abstention_rate": 0.06948228882833787,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.02666785328274954,
+        "precision": 0.9094781682641108,
+        "present_accuracy": 0.9089941458222459,
+        "recall": 0.9089941458222459,
+        "schema_valid_rate": 1.0,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.012779552715654952
+      },
+      "variant/nullable": {
+        "absent_accuracy": 0.7241379310344828,
+        "count": 28,
+        "counts": {
+          "absent_correct": 21,
+          "absent_total": 29,
+          "exact": 16,
+          "false_abstention": 6,
+          "field_correct": 57,
+          "fields": 75,
+          "fn": 10,
+          "fp": 9,
+          "json_parse_ok": 27,
+          "missed_abstention": 8,
+          "natural_stop": 28,
+          "nonrepetitive_nonempty": 28,
+          "predicted_values": 45,
+          "present_correct": 36,
+          "present_total": 46,
+          "schema_ok": 27,
+          "token_limited": 0,
+          "tp": 36,
+          "unsupported_values": 4
+        },
+        "exact": 0.5714285714285714,
+        "f1": 0.7912087912087912,
+        "false_abstention_rate": 0.13043478260869565,
+        "field_accuracy": 0.76,
+        "json_parse_rate": 0.9642857142857143,
+        "missed_abstention_rate": 0.27586206896551724,
+        "natural_stop_rate": 1.0,
+        "nonrepetitive_nonempty_rate": 1.0,
+        "output_nll": 0.08004708328683462,
+        "precision": 0.8,
+        "present_accuracy": 0.782608695652174,
+        "recall": 0.782608695652174,
+        "schema_valid_rate": 0.9642857142857143,
+        "token_limited_rate": 0.0,
+        "unsupported_value_rate": 0.08888888888888889
+      }
+    },
+    "protocol": {
+      "constrained_decoding": false,
+      "json_assembly": false,
+      "max_tokens": 256,
+      "model_calls_per_request": 1,
+      "output_repair": false,
+      "prompt_truncation": false,
+      "repetition_penalty": 1,
+      "scorer": "strict-raw-json100-v1",
+      "serialization": "native-sft-chat-v1",
+      "temperature": 0
+    },
+    "scope": "Raw model-generated JSON, source-span annotation agreement. Duplicate keys, nonfinite JSON, fences, trailing prose and schema violations fail exact correctness. Flat string arrays/nullable strings in seven SNIPS domains; observed research panels, not a new blind or external benchmark."
+  },
+  "playground": {
+    "metrics": {
+      "alias": {
+        "absent_accuracy": 0.875,
+        "count": 7,
+        "counts": {
+          "absent_correct": 7,
+          "absent_total": 8,
+          "exact": 6,
+          "false_abstention": 0,
+          "field_correct": 16,
+          "fields": 17,
+          "fn": 0,
+          "fp": 1,
+          "json_parse_ok": 7,
+          "missed_abstention": 1,
+          "natural_stop": 7,
+          "nonrepetitive_nonempty": 7,
+          "predicted_values": 10,
+          "present_correct": 9,
+          "present_total": 9,
+          "schema_ok": 7,
+          "token_limited": 0,
+          "tp": 9,
+          "unsupported_values": 0
+        },
+        "exact": 0.8571428571428571,
+        "f1": 0.9473684210526316,
+        "field_accuracy": 0.9411764705882353,
+        "json_parse": 1.0,
+        "natural_stop_rate": 1.0,
+        "precision": 0.9,
+        "present_accuracy": 1.0,
+        "recall": 1.0,
+        "schema_valid": 1.0,
+        "unsupported_value_rate": 0.0
+      },
+      "all": {
+        "absent_accuracy": 0.9056603773584906,
+        "count": 54,
+        "counts": {
+          "absent_correct": 48,
+          "absent_total": 53,
+          "exact": 42,
+          "false_abstention": 3,
+          "field_correct": 118,
+          "fields": 135,
+          "fn": 13,
+          "fp": 7,
+          "json_parse_ok": 52,
+          "missed_abstention": 5,
+          "natural_stop": 54,
+          "nonrepetitive_nonempty": 54,
+          "predicted_values": 78,
+          "present_correct": 70,
+          "present_total": 82,
+          "schema_ok": 52,
+          "token_limited": 0,
+          "tp": 71,
+          "unsupported_values": 2
+        },
+        "exact": 0.7777777777777778,
+        "f1": 0.8765432098765431,
+        "field_accuracy": 0.8740740740740741,
+        "json_parse": 0.9629629629629629,
+        "natural_stop_rate": 1.0,
+        "precision": 0.9102564102564102,
+        "present_accuracy": 0.8536585365853658,
+        "recall": 0.8452380952380952,
+        "schema_valid": 0.9629629629629629,
+        "unsupported_value_rate": 0.02564102564102564
+      },
+      "authored_diagnostics": {
+        "absent_accuracy": 1.0,
+        "count": 12,
+        "counts": {
+          "absent_correct": 11,
+          "absent_total": 11,
+          "exact": 7,
+          "false_abstention": 2,
+          "field_correct": 29,
+          "fields": 34,
+          "fn": 6,
+          "fp": 2,
+          "json_parse_ok": 12,
+          "missed_abstention": 0,
+          "natural_stop": 12,
+          "nonrepetitive_nonempty": 12,
+          "predicted_values": 21,
+          "present_correct": 18,
+          "present_total": 23,
+          "schema_ok": 12,
+          "token_limited": 0,
+          "tp": 19,
+          "unsupported_values": 1
+        },
+        "exact": 0.5833333333333334,
+        "f1": 0.8260869565217391,
+        "field_accuracy": 0.8529411764705882,
+        "json_parse": 1.0,
+        "natural_stop_rate": 1.0,
+        "precision": 0.9047619047619048,
+        "present_accuracy": 0.782608695652174,
+        "recall": 0.76,
+        "schema_valid": 1.0,
+        "unsupported_value_rate": 0.047619047619047616
+      },
+      "nullable": {
+        "absent_accuracy": 0.625,
+        "count": 7,
+        "counts": {
+          "absent_correct": 5,
+          "absent_total": 8,
+          "exact": 4,
+          "false_abstention": 0,
+          "field_correct": 12,
+          "fields": 17,
+          "fn": 2,
+          "fp": 2,
+          "json_parse_ok": 6,
+          "missed_abstention": 3,
+          "natural_stop": 7,
+          "nonrepetitive_nonempty": 7,
+          "predicted_values": 9,
+          "present_correct": 7,
+          "present_total": 9,
+          "schema_ok": 6,
+          "token_limited": 0,
+          "tp": 7,
+          "unsupported_values": 1
+        },
+        "exact": 0.5714285714285714,
+        "f1": 0.7777777777777778,
+        "field_accuracy": 0.7058823529411765,
+        "json_parse": 0.8571428571428571,
+        "natural_stop_rate": 1.0,
+        "precision": 0.7777777777777778,
+        "present_accuracy": 0.7777777777777778,
+        "recall": 0.7777777777777778,
+        "schema_valid": 0.8571428571428571,
+        "unsupported_value_rate": 0.1111111111111111
+      },
+      "observed_originals": {
+        "absent_accuracy": 0.9615384615384616,
+        "count": 28,
+        "counts": {
+          "absent_correct": 25,
+          "absent_total": 26,
+          "exact": 25,
+          "false_abstention": 1,
+          "field_correct": 61,
+          "fields": 67,
+          "fn": 5,
+          "fp": 2,
+          "json_parse_ok": 27,
+          "missed_abstention": 1,
+          "natural_stop": 28,
+          "nonrepetitive_nonempty": 28,
+          "predicted_values": 38,
+          "present_correct": 36,
+          "present_total": 41,
+          "schema_ok": 27,
+          "token_limited": 0,
+          "tp": 36,
+          "unsupported_values": 0
+        },
+        "exact": 0.8928571428571429,
+        "f1": 0.9113924050632912,
+        "field_accuracy": 0.9104477611940298,
+        "json_parse": 0.9642857142857143,
+        "natural_stop_rate": 1.0,
+        "precision": 0.9473684210526315,
+        "present_accuracy": 0.8780487804878049,
+        "recall": 0.8780487804878049,
+        "schema_valid": 0.9642857142857143,
+        "unsupported_value_rate": 0.0
+      },
+      "schema_stress": {
+        "absent_accuracy": 0.75,
+        "count": 14,
+        "counts": {
+          "absent_correct": 12,
+          "absent_total": 16,
+          "exact": 10,
+          "false_abstention": 0,
+          "field_correct": 28,
+          "fields": 34,
+          "fn": 2,
+          "fp": 3,
+          "json_parse_ok": 13,
+          "missed_abstention": 4,
+          "natural_stop": 14,
+          "nonrepetitive_nonempty": 14,
+          "predicted_values": 19,
+          "present_correct": 16,
+          "present_total": 18,
+          "schema_ok": 13,
+          "token_limited": 0,
+          "tp": 16,
+          "unsupported_values": 1
+        },
+        "exact": 0.7142857142857143,
+        "f1": 0.8648648648648649,
+        "field_accuracy": 0.8235294117647058,
+        "json_parse": 0.9285714285714286,
+        "natural_stop_rate": 1.0,
+        "precision": 0.8421052631578947,
+        "present_accuracy": 0.8888888888888888,
+        "recall": 0.8888888888888888,
+        "schema_valid": 0.9285714285714286,
+        "unsupported_value_rate": 0.05263157894736842
+      }
+    },
+    "protocol": {
+      "constrained": false,
+      "json_assembly": false,
+      "max_tokens": 256,
+      "model_calls_per_request": 1,
+      "output_repair": false,
+      "raw_text_preserved": true,
+      "repetition_penalty": 1,
+      "route": "/api/json/generate",
+      "seed": 2026,
+      "temperature": 0,
+      "top_k": 40
+    },
+    "scope": "All actual playground source/schema fixtures, with their displayed schema property ordering and instructions unchanged. Observed original DEV cases, previously authored diagnostics, and observed schema-stress views; local diagnostic, not a fresh blind benchmark. NLL not measured through generation API."
+  },
+  "historical_protocol_note": "SFT-093 results used labeled fieldwise generation plus JSON assembly; they do not describe the current raw JSON model."
+}

LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

PACKAGING_VERIFICATION.json

@@ -0,0 +1,491 @@
+{
+  "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
+  ]
+}

README.md

@@ -0,0 +1,376 @@
+---
+language:
+- en
+pipeline_tag: text-generation
+library_name: mlx
+base_model: OpenSML/OpenSML-Base-175M
+base_model_relation: finetune
+tags:
+- information-extraction
+- structured-output
+- json
+- small-language-model
+- apple-silicon
+- from-scratch
+- post-training
+- research
+datasets:
+- HuggingFaceTB/smollm-corpus
+- HuggingFaceTB/dclm-edu
+- HuggingFaceFW/finewiki
+- Avelina/python-edu-cleaned
+- rajpurkar/squad_v2
+- databricks/databricks-dolly-15k
+---
+
+# OpenSML-JSONLite-175M-Instruct
+
+**Author:** William Zebrowski · **175,611,648 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
+{
+  "type": "object",
+  "properties": {
+    "track": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      },
+      "description": "track name"
+    },
+    "artist": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      },
+      "description": "performer or musical artist"
+    },
+    "album": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      },
+      "description": "album name"
+    }
+  },
+  "required": [
+    "track",
+    "artist",
+    "album"
+  ],
+  "additionalProperties": false
+}
+```
+
+**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
+{
+  "type": "object",
+  "properties": {
+    "restaurant_name": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      },
+      "description": "restaurant name"
+    },
+    "city": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      },
+      "description": "city or town"
+    },
+    "party_size_number": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      },
+      "description": "explicit party size"
+    },
+    "timeRange": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      },
+      "description": "date, time, or time-range expression"
+    }
+  },
+  "required": [
+    "restaurant_name",
+    "city",
+    "party_size_number",
+    "timeRange"
+  ],
+  "additionalProperties": false
+}
+```
+
+**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
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "type": "object",
+  "properties": {
+    "service": {
+      "type": [
+        "string",
+        "null"
+      ],
+      "description": "music service"
+    },
+    "music_item": {
+      "type": [
+        "string",
+        "null"
+      ],
+      "description": "word specifying the requested music category, such as artist, track, song, or album"
+    },
+    "playlist": {
+      "type": [
+        "string",
+        "null"
+      ],
+      "description": "playlist name"
+    }
+  },
+  "required": [
+    "service",
+    "music_item",
+    "playlist"
+  ],
+  "additionalProperties": false
+}
+```
+
+**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.

RUN_GUIDE.md

@@ -0,0 +1,67 @@
+# Run OpenSML-JSONLite-175M-Instruct: native JSON
+
+The current weights generate the entire JSON object from source text and the
+supplied schema. No fieldwise assembler, constrained decoder or repair is used.
+The runnable package includes model.safetensors, tokenizer/config, native MLX
+model source, a source archive, 54 examples, results and verification records.
+Verified platform: Apple Silicon macOS, Python 3.11, 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 --example 29 --json
+python inference.py --text-file request.txt --schema-file schema.json
+```
+
+Default stdout is the unmodified generated text. `--json` prints a result record
+with raw `text`, parsed `value`, structural success, stop reason, token count and
+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.`
+
+```json
+{"type":"object","properties":{"track":{"type":"array","items":{"type":"string"},"description":"track name"},"artist":{"type":"array","items":{"type":"string"},"description":"performer or musical artist"},"album":{"type":"array","items":{"type":"string"},"description":"album name"}},"required":["track","artist","album"],"additionalProperties":false}
+```
+
+Actual native output:
+
+```json
+{"track":["Blue Lantern"],"artist":["Mira Sol"],"album":[]}
+```
+
+Schemas must be flat closed objects with 1–8 required properties: strings,
+nullable strings or arrays of strings. The training scope is primarily SNIPS
+fields; unfamiliar field descriptions are accepted but are not reliable by
+construction. Missing arrays use []; missing nullable strings use null. Clear
+descriptions are important. Default decoding is greedy, repetition penalty 1,
+max 256 output tokens. `--max-tokens` supports 1..256. Source + full schema +
+fixed instruction + output budget must fit 2,048 tokens; overflow rejects.
+
+The runner supplies the exact fixed instruction and native SFT serialization.
+It checks model and source-archive SHA-256 before loading. Root tokenizer files
+mirror the tokenizer inside the archive, including its integrity companions.
+After downloading and installing dependencies, generation runs offline.
+This custom architecture is not a Transformers or MLX-LM auto-loader export.
+
+```bash
+shasum -a 256 -c SHA256SUMS
+python -m zipfile -e native_bundle.zip native_bundle
+python native_bundle/scripts/verify_snapshot.py
+python native_bundle/inference.py --weights "$PWD/model.safetensors" --example 29 --json
+```
+
+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.

SHA256SUMS

@@ -0,0 +1,22 @@
+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
+619f500553367eb76f5be00a98625130ec857b9c367561334984096c029ba2a9  tokenizer_config.json
+1194292c6d906ac19e6f01c6ad2b42825143fad426a22aa8856d903776701ecb  tokenizer_manifest.json

SOURCE_NOTICE.md

@@ -0,0 +1,24 @@
+# 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.
+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.

TECHNICAL_REPORT.md

@@ -0,0 +1,314 @@
+# OpenSML-JSONLite-175M-Instruct: native JSON technical report
+
+**Author:** William Zebrowski · **Parameters:** 175,611,648 · **Date:** 2026-10-14
+
+## 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.
+
+| Stage | Inherited update | Presentations through selected artifact | Serialized positions |
+| --- | ---: | ---: | ---: |
+| Stage A pretrained | 132005 | Document stream | 17,300,111,360 lifetime pretraining positions |
+| SFT-065 grounded QA | 469 of 782 | 15,008 | 3,097,331 |
+| 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.
+
+## 3. Architecture and parameterization
+
+The same architecture is retained across pretraining and all inherited SFT stages. No new extraction head, adapter or vocabulary was introduced.
+
+| Setting | Value |
+| --- | --- |
+| Architecture | Causal decoder-only transformer |
+| Parameters | 175,611,648 |
+| Layers | 24 |
+| Hidden size | 768 |
+| Query heads | 12 |
+| KV heads | 4 |
+| Head size | 64 |
+| Attention | Grouped-query causal attention |
+| Position encoding | RoPE, base 10,000 |
+| Normalization | Pre-norm RMSNorm; query/key RMSNorm enabled |
+| Feed-forward | SwiGLU |
+| FFN hidden size | 2048 |
+| Configured MLP ratio / rounding | 4.0 / multiple 256; implementation uses 2/3 sizing |
+| Embedding / output projection | Tied weights |
+| Projection biases | Disabled |
+| Vocabulary | 32000 |
+| Context per model call | 2048 |
+| Attention implementation | MLX fast attention |
+| FFN implementation | Packed Metal SwiGLU |
+| Training CE implementation | Metal cross entropy, FP32 reductions |
+
+Implementation evidence: [`model.py`](../provenance/LINEAGE.json), [`precision.py`](../provenance/LINEAGE.json), and the final bundle's `recipe.model` metadata. Weight count and file size are consistent with BF16 parameters plus the safetensors header. GPU allocation readings are not equivalent to file size, total process RSS or system RAM requirements.
+
+The base initialization is recorded as `normal-0.02-residual-scaled-v1`. Earlier OpenSML-150M **weights** were not used to initialize the base; its frozen tokenizer was reused.
+
+## 4. Tokenizer and serialization
+
+The model uses the project's frozen 32K byte-level BPE tokenizer at `reference/opensml-150m/tokenizer/bytebpe32k_v1`. No tokenizer retraining occurs in the inherited SFT stages. Its own manifest has format `sml-v2-tokenizer-v1`; the format name is not a model-weight lineage claim.
+
+| Tokenizer artifact | SHA-256 |
+| --- | --- |
+| Tokenizer manifest / contract fingerprint | 1194292c6d906ac19e6f01c6ad2b42825143fad426a22aa8856d903776701ecb |
+| tokenizer.json | e095f0413a11fa91fbb58f37d00e4c6bcda59beb89b28f6f680db1436a7e8891 |
+| tokenizer_config.json | 619f500553367eb76f5be00a98625130ec857b9c367561334984096c029ba2a9 |
+| tokenizer corpus.json | e4c47b7b6aba08f948de7c2c7cb94c58dbd09dde2ded6ac185c60856ee686d27 |
+
+Structural token IDs:
+
+| ID | Token | Meaning |
+| --- | --- | --- |
+| 0 | `<\|pad\|>` | Padding |
+| 1 | `<\|doc_end\|>` | Document/end-of-sequence marker |
+| 2 | `<\|turn_start\|>` | Role/turn header |
+| 3 | `<\|turn_end\|>` | End of a conversational-format turn |
+
+`pretraining/sft/data.py` constructs explicit IDs rather than assuming a generic Hugging Face chat template. A user turn is `turn_start + encode("user\n") + content + turn_end`. An assistant target has the corresponding assistant header, response and turn-end marker; training additionally appends document-end. Generation ends its prompt at the assistant header. User text and role headers are masked from the supervised loss. Assistant content, assistant turn-end and the appended document-end are supervised.
+
+Ordinary text must round-trip through the tokenizer and cannot emit reserved structural IDs. Special-token-looking content can be rejected by the tokenizer; the API does not promise unrestricted literal handling of all structural strings. Prompt/output limits are checked without truncating source text.
+
+## 5. Pretraining corpus and curation
+
+The selected base is the best retained checkpoint at update 132005 after **17,300,111,360 processed token positions**. The original budget was 20B; a continuation was configured toward 30B, but this inherited checkpoint is not a completed 20B or 30B artifact.
+
+Source weights below are intended processed-token sampling weights, not measured unique-document shares. No claim of 17.3B globally unique corpus tokens is made.
+
+| 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.
+
+The exact corpus specification is [`configs/corpus_20b.json`](../provenance/LINEAGE.json). Training preserves source text and code indentation. NFC/whitespace normalization is used for identity hashing and partitioning, not to rewrite model input.
+
+The content-split policy hashes identity into 10,000 buckets: 97% training, 1% validation, 1% final and 1% tokenizer development. Identity checks do not establish that a supervised benchmark article never appeared elsewhere in pretraining web data.
+
+Recorded stream settings are a 256-document accepted-document shuffle buffer, bounded 100,000-document deduplication window, 4 prefetched batches and up to 5 retry attempts. The corpus JSON's generic 10,000 shuffle value is not the active stream override. Filtering rejects documents below 200 characters, above 200,000 UTF-8 bytes, with null characters, excess replacement characters or an excessive repeated-line pattern. Source exhaustion fails rather than silently recycling. Deduplication is bounded; it is not global semantic or fuzzy deduplication.
+
+## 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 |
+| --- | --- | --- | --- | --- |
+| 1 | M5 Ultra | 96 GB | 7 | 16 GiB |
+| 2 | M3 Ultra | 96 GB | 3 | 16 GiB |
+| 3 | M4 Max | 36 GB | 2 | 4 GiB |
+| 4 | M4 Max | 36 GB | 2 | 4 GiB |
+| 5 | M4 Max | 36 GB | 2 | 4 GiB |
+
+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.
+
+Ethernet supports SSH orchestration; the training ring uses verified Thunderbolt/RDMA links. Sequential microbatch materialization, FP32 gradient accumulation and custom Metal paths were used. Public documentation must not include private SSH identities, addresses, access credentials or machine serial numbers.
+
+| Pretraining optimizer/schedule | Recorded setting |
+| --- | --- |
+| Optimizer | AdamW |
+| Nominal peak LR | 3e-4 |
+| Original warmup | 200M tokens, 1% of the 20B budget |
+| Base schedule | Warmup/stable interval/cosine; continuation stretches nominal decay to 24B–30B |
+| Adaptive controller | Factor 0.8; patience 12 evaluations; cooldown 10; minimum delta 0.002 |
+| Controller LR floor | 3e-5 |
+| Effective cap at selected 132005 | 3.221225472e-5 |
+| Recorded controller reductions / evaluations | 9 / 242 |
+| Betas | (0.9,0.95) |
+| Epsilon | 1e-8 |
+| Weight decay | 0.1 |
+| Gradient clipping | 1.0 |
+| Seed | 7337 |
+| Loss | Next-token cross entropy; FP32 reductions |
+
+The nominal 3e-4 schedule is not the effective late-run learning rate: the retained controller cap bounds it. The record does not justify a claim that the run trained at 3e-4 for all 17.3B positions.
+
+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
+
+The 175.6M-parameter base was trained from random initialization using next-token prediction across the educational web, Wikipedia, Cosmopedia and Python mixture. Five Apple Silicon Macs carried out distributed pretraining over Thunderbolt/RDMA. Validation was measured every 50M tokens. The curves show a steep early loss reduction followed by slower improvement; the published base was selected using held-out validation.
+
+![Stage A pretraining training loss, EMA and held-out validation loss](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.*
+
+
+## 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.

THIRD_PARTY_NOTICES.md

@@ -0,0 +1,20 @@
+# Third-party notices
+
+The root Apache-2.0 license covers author-controlled source contributions. Installed dependencies and training materials retain their upstream terms. Source, datasets and trained weights have separate licensing status.
+
+| Source | Role | Recorded declaration / attribution |
+| --- | --- | --- |
+| [SmolLM corpus](https://huggingface.co/datasets/HuggingFaceTB/smollm-corpus) | FineWeb-Edu and Cosmopedia pretraining | ODC-BY source card; underlying material retains its own notices. Pinned revisions are in the corpus file. |
+| [DCLM-Edu](https://huggingface.co/datasets/HuggingFaceTB/dclm-edu) | Pretraining | CC-BY-4.0 source declaration. |
+| [FineWiki](https://huggingface.co/datasets/HuggingFaceFW/finewiki) | English pretraining | CC-BY-SA-4.0/GFDL source declaration. |
+| [Python-Edu mirror](https://huggingface.co/datasets/Avelina/python-edu-cleaned) | Educational Python pretraining | Local metadata records ODC-BY; underlying code repositories and The Stack v2/SWH provenance retain their conditions. The pinned mirror card was not accessible during the report's external audit. |
+| [SQuAD 2.0](https://rajpurkar.github.io/SQuAD-explorer/) | Grounded QA SFT | Pranav Rajpurkar, Robin Jia and Percy Liang; CC-BY-SA-4.0. Original answer offsets were checked and a curated subset retained. Raw training corpus is not distributed. |
+| [Dolly 15k](https://huggingface.co/datasets/databricks/databricks-dolly-15k) | Selected closed QA SFT | Databricks dataset contributors; CC-BY-SA-3.0. Raw training corpus is not distributed. |
+| [SNIPS NLU benchmark](https://github.com/sonos/nlu-benchmark/tree/b86ac7f1577868c42158d0dec77db50956046696) | Exact field extraction, native JSON SFT and adapted held-out evaluation | SNIPS/Sonos contributors; CC0-1.0. Original chunk annotations transformed into requested-field views; [included license](licenses/SNIPS-CC0.txt). |
+| OpenSML authored fictional facts | Controlled QA examples | Programmatically authored and checked; not independent human certification. |
+
+MLX, NumPy, Tokenizers and jsonschema are installed dependencies, not vendored distributions in this archive. Their upstream distributions supply their license files. This repository includes author-controlled MLX model/kernel and orchestration source from the OpenSML workspace; the original first release's public source repository is [opensml-150m](https://github.com/williamzebrowskI/opensml-150m).
+
+No IFEval, Paraloq, SOB, ARC, PIQA, HellaSwag or other experiment's scores are advertised as evaluation of these weights. Older research helpers present in dependency packages do not change the inherited training lineage.
+
+For actual source-file pins and identity-preserving adaptations, see the technical report and `provenance/`. These declarations are attribution records, not a blanket commercial-use or model-weight redistribution determination. The public weight license remains unassigned.

checkpoint_metadata.json

@@ -0,0 +1,174 @@
+{
+  "extraction_sft": {
+    "contract": {
+      "code": {
+        "pretraining/extract_068/data.py": "497b0b12c8896c2045f470c1c2de0330472663f1fd5b67746734a3f7907ee971",
+        "pretraining/extract_068/evaluate.py": "a7ff05b0d2383421e83d0247e613017715485b3a5fffdcc7240b90387a756c06",
+        "pretraining/extract_068/runtime.py": "7b97df4835d03210c77da36f53340aa83d5915b9cfa7aae968aef6961422c2f1",
+        "pretraining/extract_068/train.py": "f44af1b00a7e36c3dd4c42c30dc63b73bc0d5917495f1985cc6bc3c5b902b6c5",
+        "pretraining/extraction_001/__init__.py": "f1e824c48131bc441020c864929875bfa44daebba75b15181cc401f7b73ae8a2",
+        "pretraining/extraction_001/data.py": "59843f3beca3aecbe8e905af8c969fa9fd6a42554da8d2b1054d027fb771a659",
+        "pretraining/extraction_001/evaluate.py": "b22da13e22b5ebffeb217200e5ff431773d85e6ae394045e96c97ed1bc7e7c54",
+        "pretraining/extraction_001/train.py": "071ef9293bb825753b47b3704e1df1e69f33b792c45c5748f226e40c6c054e26",
+        "pretraining/json_evidence_binding/__init__.py": "357162457d07712f819559c1f1a6769323941355531cb6a04f657ace44d5c52c",
+        "pretraining/json_evidence_binding/data.py": "fd828fcaa202085ea38a4a927e31f7bfc09e19a8965cd170f095408744e3a8a0",
+        "pretraining/json_evidence_binding/evaluate.py": "25fc997a4035ab44f6e5ac47988abc38886efb2c23d8cc270f014e8c0a04cc71",
+        "pretraining/json_evidence_binding/objective.py": "5bcfe1d57e445ca79a82daf81a8d664bb1c8178aaf3ff4189665e205a95537aa",
+        "pretraining/json_evidence_binding/order.py": "37759caee925d4eb2a6e8273075524090516017c199fbc9ebe75edee9b1a7cd6",
+        "pretraining/json_evidence_binding/runtime.py": "4d6695edd7eccfd26ca31a282e76ae663fe149b86f6fc42e8edc4e7625ca15ec",
+        "pretraining/json_evidence_binding/storage.py": "a93d3af323904983f016870865e90ee89a02df4c66d15e502ef9a2ebfff99424",
+        "pretraining/json_evidence_binding/train.py": "2e20b5a09501145a68500994eed4a8e5134148d805dafd93726ea09afca23ebf",
+        "pretraining/json_sft/__init__.py": "aa9dcf15611647559bbc4ab1d8dfa3746cf731f30e4b5fd35bf53a1ed00bea95",
+        "pretraining/json_sft/adapters.py": "09dd78767c9f7e0dbaa0184e64d0011afadd7a326986766026dfcc64106ab8d9",
+        "pretraining/json_sft/benchmark.py": "75ae0df6b0d6676805d6c1ef3b7b73893f473df001dcc2a542f31fb074dd3601",
+        "pretraining/json_sft/business.py": "5c31088adce5f85ce8980d159fa0f9c39a0e1d9a7725560838a2852ac5cfbff4",
+        "pretraining/json_sft/constrained.py": "dc21aa142fe525f3559cd5638e46ef287555efba8239a06072a092fc10700cf5",
+        "pretraining/json_sft/contrasts.py": "e504c5f879011963b82728d73fe334585a640d419c4e3d47d919a4cc89f396fd",
+        "pretraining/json_sft/corpus.py": "37d7b74e0478853903953934b5d2006fd059e1ca2a613cc69f6788e9f34d0e6d",
+        "pretraining/json_sft/decomposed.py": "46f863e628bcc5fedfd6f34f7c759d4b0998091d8252b48729027ce9ef2a01b9",
+        "pretraining/json_sft/diagnostics.py": "92f750807aebb9cb38154c4599d468850ce25ef86496de8e660a77e538e08f03",
+        "pretraining/json_sft/evaluate.py": "2c3076f69ef2d7721f25a79618563d102484bedbf2912f684b644b9135d25d17",
+        "pretraining/json_sft/expansion.py": "231274b7e6d486d39bf962ff310ad4557ccfd84f8cac329cf57ed5890aed7cc7",
+        "pretraining/json_sft/grounded.py": "f668379c5f904ee36c267f6897f810cdb04f80e19a8fd5fdfb791b31f902e8c9",
+        "pretraining/json_sft/narrative.py": "9d507dbc7c2eff7a4b77064393dc65e65c6e3a1866b4375df3d8b66c755d5be1",
+        "pretraining/json_sft/ordered.py": "0ac1b6dd3139425c06094cf5e185ffab5389a377ecaf58b6320978908df4b7ef",
+        "pretraining/json_sft/ordered_serialization.py": "14c23f875b45d17d8d07cb106afc29a6b72cb9b7bb5a393ee0ee55bf84b5aa1a",
+        "pretraining/json_sft/plain.py": "64dd1060d5a674ef04bba3ea3965b7e89ab4faa58bbc3c437816ed2023023843",
+        "pretraining/json_sft/precise.py": "7696328b55290548f05562a9562ea898301a992b4161000f5f2dd7d1bbaf621a",
+        "pretraining/json_sft/reader.py": "ff288fddabd0525f8fc6a8f2a6aeaedf309cfb954448d8c6aa57e122e8628d6e",
+        "pretraining/json_sft/relational.py": "d2e5d3f850b5b543876fafac8b3d57c3ea7c6983437ba6744026ed86701e929e",
+        "pretraining/json_sft/schedule.py": "50ec8a06b0f881714451501b6325a0a64c7cbff159b57f1214f0a3a6ac83014c",
+        "pretraining/json_sft/serialization.py": "d04338a44b569c3f38ed42b0e1135738c1c119387134945335d51bb354c2213f",
+        "pretraining/json_sft/serving.py": "d533a59791bd2516dd1ec4cd2493c554161c0fe1e3603d55fa12631491cddbf8",
+        "pretraining/json_sft/slots.py": "8db7eb31a53899b9e4306af97fad6219ff6c8054a0df67ceffe0dbe927e5df24",
+        "pretraining/json_sft/train.py": "abe0613af70c559dca777cc1fc3c905be0f9583fb6aa30442881da8d0a95f316",
+        "pretraining/json_sft/transfer.py": "ff5bc65fae8da93164f830cff5a20fcff42132bcbf02a94de672af9653cf340d",
+        "pretraining/json_sft/unquoted.py": "969263db40f1a5561acbafdfb21dbb74ce85dd394c2bc9309cd208f61c977289",
+        "pretraining/json_sft/value_weights.py": "043cde3b06edcbc0bd332eb8ace4257e1864af77c239204c120261b95696598b",
+        "pretraining/native_json_100/__init__.py": "8d4b1be2f7d75122c982cd881c19a41b1d39594c5e30fbbe2bc87fcccf9c0242",
+        "pretraining/native_json_100/data.py": "594a83b846c97836be009e61d9fcf9a1b4ceb0d0f9bb491bbb5b56f02a433edf",
+        "pretraining/native_json_100/evaluate.py": "3c03d63aeb85ee85d857f219ee3aa4c190f34532ab6643cb487ca8e41d9779c3",
+        "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",
+        "pretraining/qa_064/train.py": "9400aa40b0f5b332334e11fc8b62f5d06464fe1188603bb65585302461d3bf19",
+        "pretraining/qa_065/controlled.py": "55799050001b55a642705064bf79ce45625eca81ebfe77d7f70cba8435726f51",
+        "pretraining/qa_065/data.py": "1e4878ce7bd9e48d8d5c79eafac31093eaf7cdcdd4a6aeb5f9b6519967e47e57",
+        "pretraining/qa_065/evaluate.py": "f8e4a209a7cdb42cded3cb5ffdb2aabfa0d12cc13c10c2f40f33907c532f74e5",
+        "pretraining/qa_065/runtime.py": "7b97df4835d03210c77da36f53340aa83d5915b9cfa7aae968aef6961422c2f1",
+        "pretraining/qa_065/train.py": "52b11949018bc87c6c80dc3bee7571b2ca95967ebb47fe450a5f5552c46d4a16",
+        "pretraining/qa_066/controlled.py": "fc101c488049fd836e53422a72d13d8d02f253153726e9e4d5eda0eb19ab9baa",
+        "pretraining/qa_066/data.py": "4b89e4107bd7076270151989871805af4162d76a299e72d40881fdc26750015a",
+        "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",
+        "pretraining/sml_v1/__init__.py": "7e9e02294e03e6b8a12fd202ab288f2f9b0073bb11b8cb2a5fa82211b0d1dfa4",
+        "pretraining/sml_v1/batch_transition.py": "e6778ba8a8215f0b3c20386e768ea3e15b5939e7919935046689733eb96eba28",
+        "pretraining/sml_v1/checkpoint_bundle.py": "73cbf0c17cf74a44bb7c74a8f9e5bb63671057a1f664a6a4af0e0777b2c8f1a4",
+        "pretraining/sml_v1/checkpoint_metadata.py": "0aaa7019ab53fd8b5e6b0e397f970026b4e8d38e12bf5d5f0fcb7f686de9b82c",
+        "pretraining/sml_v1/common.py": "ec4dda551cbe45cbd30d64c7c82379a8849ce02a3ad1a597a293844b5dc86351",
+        "pretraining/sml_v1/continuation.py": "3c5248646c27bcdf6bd4010742fe6e6d0aa9f059b543e183b60e00e3dbe034a3",
+        "pretraining/sml_v1/corpus.py": "be77769180342f6a8b1d5619e5b3b64f20ca40d3424219168cccf4789bb02038",
+        "pretraining/sml_v1/data.py": "411284cda81b07f10bc624ebd8f7fe9debd2911e950971f96b204dfd1cee6a6d",
+        "pretraining/sml_v1/data_review.py": "a71eabcc526344b6502cae999020da63c950b2d80c13077629bb86c8757a18ed",
+        "pretraining/sml_v1/diagnostic_eval.py": "317c05d4d17239e62d1316797892288f5e4eae467b3e66a4ca4f15d189960a9b",
+        "pretraining/sml_v1/fast_loss.py": "f99edeedb6cb328ef728d05643f65ea20371437f7e5806009ae3ad0885bafbf1",
+        "pretraining/sml_v1/fast_swiglu.py": "59d1adb31400e3c288914b4c0cc8a21681a73ca3dfa7c722ea4e13d21dc38528",
+        "pretraining/sml_v1/inference.py": "4255b93bbc18d120318d83fb26b2fda1fe16d19ca1423250c77188940cc4f008",
+        "pretraining/sml_v1/jaccl_control.py": "a3dfbd61de43fda14daf4bbd921ccb382ea8af2c8342457fe16d1ad0908e88fe",
+        "pretraining/sml_v1/launch.py": "41e6a98d9b41dc844063fc28ce81a08d83b919306b9daf0716b76a0bc21365fc",
+        "pretraining/sml_v1/lr_scheduler.py": "e141132efbd5f1beab4b187d4eab15efd46fcf9710d6a2ba7c882f62362a77cf",
+        "pretraining/sml_v1/lr_trial.py": "2f643abaa6ac3f86e47b37632df4d25fa34d0b25e77b59ffecc0290b2385e7dc",
+        "pretraining/sml_v1/model.py": "1eee6f8afe78767ff6b9c604ccf2328c98b24bc94dbad6291a0ad699dd7345fe",
+        "pretraining/sml_v1/playground_publish.py": "bffd8f8c5a74a0bf9c3ca070e6511cc1ab8ceb8343c0092a3d711b4fb9ece21c",
+        "pretraining/sml_v1/precision.py": "b9f535fca724153703ed57a300eceb5b34a8f8381fc65bf5c4c5a0f7e066303c",
+        "pretraining/sml_v1/prepare_data.py": "3d78ed45dbab52a671d889fe937a210a832e55426bbdc4f905ddbd202ccf2336",
+        "pretraining/sml_v1/pretrain.py": "456a498d50e991cd83363588392b91e58a8bdeec7ceaf9950671925b6a86a46b",
+        "pretraining/sml_v1/recipe.py": "d354731910ac16868b21717a089fce5199a898dec39c6561b3d10c89bfa8a2d1",
+        "pretraining/sml_v1/research_stage_b.py": "5234bc6b824e4b07f63dda48e5353bbdb5f5abac4c98c299b642d1815ce85d94",
+        "pretraining/sml_v1/sample_prompts.py": "1173c8c94dc11c6b8e446c26804a543b68e366986f01d249384082629dca369b",
+        "pretraining/sml_v1/stable_schedule.py": "631451a45b020120610b7116e8e63276ba5043011359fbd13e49e63ec8e2b02b",
+        "pretraining/sml_v1/stage_b.py": "f80f065f6892eea951e94a7ff6925d5a1bff9419a7384e0756cf2faeb92e4775",
+        "pretraining/sml_v1/stage_b_data.py": "c065be118b0c6b8948a15d4b81e7fd7371f4e699ea3c49d34fa0abde258e0505",
+        "pretraining/sml_v1/stage_b_eval.py": "943fd969994e896faff14481b94eb121d23b6fef00244874d5d5361df6bd8cce",
+        "pretraining/sml_v1/stream.py": "70e36e12d33547bfce505bd0302605e3d66c9ee65fc0facb26ccf5341bde8b4b",
+        "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": {
+        "jsonschema": "4.26.0",
+        "mlx": "0.32.3.dev20260920+7241f12",
+        "numpy": "2.4.2",
+        "tokenizers": "0.22.2"
+      }
+    },
+    "order_hash": "76518047ebfc78beb886d78322635ee275acdb6ad479a5d722bb871be3c89961",
+    "records": 15000,
+    "serialization": "native-sft-chat-v1",
+    "total_steps": 469
+  },
+  "recipe": {
+    "model": {
+      "attention_impl": "fast",
+      "bias": false,
+      "ce_impl": "metal",
+      "d_model": 768,
+      "ffn_impl": "packed-metal",
+      "loss_dtype": "float32",
+      "max_seq_len": 2048,
+      "mlp_multiple_of": 256,
+      "mlp_ratio": 4.0,
+      "n_heads": 12,
+      "n_kv_heads": 4,
+      "n_layers": 24,
+      "qk_norm": true,
+      "rope_base": 10000.0,
+      "vocab_size": 32000
+    }
+  },
+  "step": 469,
+  "tokens": 4634126
+}

checkpoint_provenance.json

@@ -0,0 +1,58 @@
+{
+  "model": "OpenSML-JSONLite-175M",
+  "stages": [
+    {
+      "stage": "stage-a",
+      "step": 132005,
+      "tokens": 17300111360,
+      "weight_sha256": "5151567e266244c9e32fd6e01775192633abe227320269ecdefe28ce43e9ccc3",
+      "weight_bytes": 351242342,
+      "original_metadata_sha256": "15a0e3b511b97503a611c0da838a7fde1c06ebed34c7103f7cda1782cc6378a4",
+      "published_metadata_sha256": "15a0e3b511b97503a611c0da838a7fde1c06ebed34c7103f7cda1782cc6378a4"
+    },
+    {
+      "stage": "sft-065",
+      "step": 469,
+      "tokens": 3097331,
+      "weight_sha256": "0991b805a878c15733c909ef7f4127baa2b8866f8ca7d7db232ad39434f80bc4",
+      "weight_bytes": 351242342,
+      "original_metadata_sha256": "f245e46ecfbbc499d7727c556ea889a7ac77f619d8d815f65267100ddb38e991",
+      "published_metadata_sha256": "1172eb192546ae0d4d395943be0628c8de0fa2d9a4de1243eec8e73fef2d5f44"
+    },
+    {
+      "stage": "sft-068",
+      "step": 379,
+      "tokens": 1842593,
+      "weight_sha256": "18f0f2b8f57905ac758f392d0eb83e51e508da6a955b9f7991338e690205df86",
+      "weight_bytes": 351242342,
+      "original_metadata_sha256": "49acc0640e80749fb143e9a4d5024d2ffba7e79385314a792a46cc23f02283ac",
+      "published_metadata_sha256": "192399fe9cbb09ff0a2db1b21d87193968cc766f177dc9ad3975f289e8121989"
+    },
+    {
+      "stage": "sft-093",
+      "step": 160,
+      "tokens": 706399,
+      "weight_sha256": "d552c1380c71c7c0569e9d692ec9c46c5c94bb87a7b024e56df0f0e36df4d798",
+      "weight_bytes": 351242342,
+      "original_metadata_sha256": "32a380899c455449db066ef317de9dd8e80f6977bfb42a6fe03b67c5f23963f8",
+      "published_metadata_sha256": "32a380899c455449db066ef317de9dd8e80f6977bfb42a6fe03b67c5f23963f8"
+    },
+    {
+      "stage": "sft-100",
+      "step": 469,
+      "tokens": 4634126,
+      "records": 15000,
+      "weight_sha256": "656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4",
+      "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"
+}

config.json

@@ -0,0 +1,34 @@
+{
+  "model_name": "OpenSML-JSONLite-175M",
+  "parameters": 175611648,
+  "model": {
+    "attention_impl": "fast",
+    "bias": false,
+    "ce_impl": "metal",
+    "d_model": 768,
+    "ffn_impl": "packed-metal",
+    "loss_dtype": "float32",
+    "max_seq_len": 2048,
+    "mlp_multiple_of": 256,
+    "mlp_ratio": 4.0,
+    "n_heads": 12,
+    "n_kv_heads": 4,
+    "n_layers": 24,
+    "qk_norm": true,
+    "rope_base": 10000.0,
+    "vocab_size": 32000
+  },
+  "weight_sha256": "656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4",
+  "serialization": "native-sft-chat-v1",
+  "output_mode": "native-json",
+  "context_limit": 2048,
+  "weight_format": "BF16 safetensors",
+  "supported_runtime": "native MLX; not a Transformers or mlx-lm auto-loading export",
+  "max_output_tokens": 256,
+  "model_calls_per_request": 1,
+  "json_assembly": false,
+  "output_repair": false,
+  "constrained_decoding": false,
+  "selected_stage": "sft-100",
+  "selected_update": 469
+}

inference.py

@@ -0,0 +1,36 @@
+#!/usr/bin/env python3
+"""Run JSONLite offline with the bundled native MLX implementation.
+
+Defaults to model.safetensors beside this script. The verified source bundle
+includes the exact model architecture, tokenizer, native JSON prompt/decoder and examples.
+"""
+from pathlib import Path
+import hashlib
+import runpy
+import sys
+import tempfile
+import zipfile
+
+ROOT = Path(__file__).resolve().parent
+BUNDLE_SHA256 = "4ea6280d2e60da34e59757a4d6ab95aaf7811050c68c047cf5b98d2f001e9b78"
+
+def main():
+    archive = ROOT / "native_bundle.zip"
+    if not archive.is_file():
+        raise SystemExit("Missing native_bundle.zip; download the complete repository.")
+    if hashlib.sha256(archive.read_bytes()).hexdigest() != BUNDLE_SHA256:
+        raise SystemExit("Native source bundle failed its integrity check.")
+    if not any(arg == "--weights" or arg.startswith("--weights=") for arg in sys.argv[1:]):
+        sys.argv.extend(["--weights", str(ROOT / "model.safetensors")])
+    with tempfile.TemporaryDirectory(prefix="jsonlite-runtime-") as directory:
+        runtime = Path(directory)
+        with zipfile.ZipFile(archive) as bundle:
+            for member in bundle.infolist():
+                path = Path(member.filename)
+                if path.is_absolute() or ".." in path.parts or (member.external_attr >> 16) & 0o170000 == 0o120000:
+                    raise SystemExit("Unsafe source-bundle entry.")
+            bundle.extractall(runtime)
+        runpy.run_path(str(runtime / "inference.py"), run_name="__main__")
+
+if __name__ == "__main__":
+    main()

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:656f45015aab548465d93bed0b57e687497163741885da7364872851fa17e9a4
+size 351242342

native_bundle.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ea6280d2e60da34e59757a4d6ab95aaf7811050c68c047cf5b98d2f001e9b78
+size 5607781

requirements.txt

@@ -0,0 +1,4 @@
+mlx==0.32.3
+numpy==2.4.2
+tokenizers==0.22.2
+jsonschema==4.26.0

tokenizer_config.json

@@ -0,0 +1,8 @@
+{
+  "add_bos_token": false,
+  "add_eos_token": false,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|doc_end|>",
+  "pad_token": "<|pad|>",
+  "tokenizer_class": "PreTrainedTokenizerFast"
+}

tokenizer_manifest.json

@@ -0,0 +1,31 @@
+{
+  "boundary_policy": "raw-text-then-one-document-EOS-v1",
+  "corpus_fingerprint": "a89f69d6f63cfa4385da233693a445404234a18194170fe1d2b4a74bf501f5d6",
+  "format": "sml-v2-tokenizer-v1",
+  "recipe": {
+    "corpus": "a89f69d6f63cfa4385da233693a445404234a18194170fe1d2b4a74bf501f5d6",
+    "min_frequency": 2,
+    "sample_manifest": "30ead6f502f11e99b479e3711a81ce7198787408d0700153615fc3255b244f46",
+    "specials": [
+      "<|pad|>",
+      "<|doc_end|>",
+      "<|turn_start|>",
+      "<|turn_end|>"
+    ],
+    "tokenizers": "0.22.2",
+    "version": 1,
+    "vocab_size": 32000
+  },
+  "sha256": {
+    "corpus.json": "e4c47b7b6aba08f948de7c2c7cb94c58dbd09dde2ded6ac185c60856ee686d27",
+    "tokenizer.json": "e095f0413a11fa91fbb58f37d00e4c6bcda59beb89b28f6f680db1436a7e8891",
+    "tokenizer_config.json": "619f500553367eb76f5be00a98625130ec857b9c367561334984096c029ba2a9"
+  },
+  "special_ids": {
+    "eos": 1,
+    "pad": 0,
+    "turn_end": 3,
+    "turn_start": 2
+  },
+  "vocab_size": 32000
+}