publish fts-specgen-dataset: 15 files (134.3 MiB)

.gitattributes

@@ -58,3 +58,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+negative_control/mutants.jsonl filter=lfs diff=lfs merge=lfs -text
+hf/negative_control.jsonl filter=lfs diff=lfs merge=lfs -text
+data/train.jsonl filter=lfs diff=lfs merge=lfs -text
+hf/train.jsonl filter=lfs diff=lfs merge=lfs -text

MANIFEST.json

@@ -0,0 +1,78 @@
+{
+  "generated_at": "2026-08-07T09:09:59+00:00",
+  "repo_id": "digitable-lol/fts-specgen-dataset",
+  "repo_type": "dataset",
+  "note": "sha256 of every file as published. The hash is the record, not the filename: a repository is the same artefact as this manifest only if every digest matches.",
+  "files": [
+    {
+      "path_in_repo": "README.md",
+      "sha256": "a05446f4c157bfd1398a5b7e5883efcd80cba3f08cf4136456b6b14f83056150",
+      "size_bytes": 26060
+    },
+    {
+      "path_in_repo": "data/holdout.jsonl",
+      "sha256": "4a73ec2a6c402af91e7a130b5af3ac12efe8106432e588dd415ccb9bd1ba5e1c",
+      "size_bytes": 7270701
+    },
+    {
+      "path_in_repo": "data/train.jsonl",
+      "sha256": "d50c09795840a8548010f3ecf7ccc460b3c1be60ee8ba4c071a775ff0abdacc2",
+      "size_bytes": 48929988
+    },
+    {
+      "path_in_repo": "fts.gbnf",
+      "sha256": "d8b3fc451c4e6339c1c3c3ef6a1f8533b870ea20c6c05573b6c28d8bb8a2573e",
+      "size_bytes": 6765
+    },
+    {
+      "path_in_repo": "hf/holdout.jsonl",
+      "sha256": "93b80ff1b68927baa1a0db692f545612c30832b1d3033f111142f5790fb877f5",
+      "size_bytes": 7823602
+    },
+    {
+      "path_in_repo": "hf/negative_control.jsonl",
+      "sha256": "b9808a3afa16f2ed6beb67479f1d5f2d4a4843dd22a11f73fc3fda0b0f61e7ee",
+      "size_bytes": 12016316
+    },
+    {
+      "path_in_repo": "hf/train.jsonl",
+      "sha256": "7f645f250af1eca9cc9a3b01065dc4deb287ad5dcbc39a73a395c05c8f703910",
+      "size_bytes": 52279316
+    },
+    {
+      "path_in_repo": "negative_control/gate_sensitivity.json",
+      "sha256": "9abe437ff59bbdf127b13dbecc104d170999c0bbcfdef33f88dfbbdd981a193e",
+      "size_bytes": 2264
+    },
+    {
+      "path_in_repo": "negative_control/mutants.jsonl",
+      "sha256": "af893c1c07d43e1c91f9b61b2a350ce1f347d03d6ca02e77de12bf94c4334e4c",
+      "size_bytes": 11066486
+    },
+    {
+      "path_in_repo": "negative_control/verdicts.jsonl",
+      "sha256": "3efe28177d11b2aa9e79e5d1ef4820158823dcff190442a5d3f7d6ebfb1aee23",
+      "size_bytes": 1430563
+    },
+    {
+      "path_in_repo": "reports/coverage_vs_corpus.json",
+      "sha256": "1104e2c43954cc6dd6bb8f90d6562b67880415209f2b313ca0786ca710401b70",
+      "size_bytes": 7034
+    },
+    {
+      "path_in_repo": "reports/grammar_conformance.json",
+      "sha256": "af31ab5cd4a386fd4a103b5988212866675363b58cdf0d43dccc5b46f1ecab94",
+      "size_bytes": 126
+    },
+    {
+      "path_in_repo": "reports/split.json",
+      "sha256": "587333373bfe4a1452abdb0c8d6edb9b2580c35b1499c2f3e8cf459accb171bd",
+      "size_bytes": 3637
+    },
+    {
+      "path_in_repo": "reports/stats.json",
+      "sha256": "53d05cc3e6a5fa0a7aeae9c3311802f3ec7d88af4902fc3fe954a0987115c1fe",
+      "size_bytes": 9327
+    }
+  ]
+}

README.md

@@ -0,0 +1,445 @@
+---
+license: bsd-2-clause
+language:
+  - ru
+task_categories:
+  - text-generation
+tags:
+  - formal-specification
+  - executable-specification
+  - fts
+  - russian
+  - synthetic
+  - verified
+  - compiler-checked
+  - negative-control
+size_categories:
+  - 10K<n<100K
+configs:
+  - config_name: default
+    data_files:
+      - split: train
+        path: hf/train.jsonl
+      - split: holdout
+        path: hf/holdout.jsonl
+  - config_name: negative_control
+    data_files:
+      - split: train
+        path: hf/negative_control.jsonl
+---
+
+# fts-specgen-dataset
+
+**10 500 pairs of "task in ordinary Russian → executable FTS specification".**
+9 000 train, 1 500 holdout. **Every single document was run through the real FTS compiler**
+— parsed, type checked, its examples executed, its theorem proved and certified — and only
+what passed the whole gate is in these files.
+
+It also ships the thing that makes that sentence worth anything: **a negative control of
+4 400 deliberately corrupted documents**, eleven kinds of corruption, with the gate's
+verdict on every one. The gate catches **91.6 %**, and the card tells you which 8.4 % it
+misses and why.
+
+This is what trained [`digitable-lol/specgen-qwen3-1.7b`](https://huggingface.co/digitable-lol/specgen-qwen3-1.7b).
+
+---
+
+## 1. Read this first
+
+### 1.1 "Passed the compiler" is a claim about form, not about meaning
+
+The gate proves that a document **compiles, type checks, executes its own examples, and
+proves its own theorem**. It does not prove that the document means what the task text
+asked for.
+
+**Semantic fidelity was never measured.** The generator builds the task text and the
+document from the same underlying structure, so a divergence can only arise as a bug in
+`tools/task.py` — but the rate of such bugs is unknown, and "unknown" is the honest word.
+If you need a corpus where the natural-language request is guaranteed to describe the
+formal artefact, this is not verified to be one.
+
+### 1.2 A hole in this dataset, found by the model that trained on it
+
+**Not one of the 9 000 training documents contains a utility and a theorem at the same
+time.**
+
+| training documents | count |
+|---|---:|
+| utility, no theorem | 6 451 |
+| theorem, no utility | 1 185 |
+| neither | 1 364 |
+| **both** | **0** |
+
+The holdout does contain the combination — template `CMB-UTL-THM`, 129 documents — because
+that template was assigned to the holdout side of the split. The consequence showed up
+downstream, in the model: of the 127 `CMB-UTL-THM` documents it produced that compiled
+cleanly, **0 contained the theorem the task asked for**, while on the other 324
+theorem-bearing holdout documents it scored 324/324.
+
+This is a defect of the data, not of the model, and it is a good demonstration of § 1.1:
+the compiler accepts a specification with no theorem, because a specification with no
+theorem is formally valid. Nothing in the gate could have caught it. It was found by
+comparing generated documents against their reference structure — a check this dataset's
+own construction never had to perform.
+
+If you train on this corpus, the model you get will not learn to combine a utility with a
+theorem in one document, because it will never have seen it done.
+
+### 1.3 Russian only, and one specific language
+
+Every task text is Russian and every document uses the **Russian surface syntax** of FTS.
+The English surface of the language is not covered at all. The documents are also specific
+to FTS as implemented by [`@digitable-lol/fts`](https://github.com/digitable-lol/flang) —
+this is not a general formal-methods corpus.
+
+## 2. What is in it
+
+| | |
+|---|---|
+| documents | **10 500** (9 000 train / 1 500 holdout) |
+| passed the gate | **10 500 of 10 500** generated candidates |
+| rejected in the final run | 0 |
+| subject areas | **62** (50 train, 12 holdout) |
+| structural templates | **33** (25 train, 8 holdout) |
+| language constructs tracked | **74 — all 74 present in both train and holdout** |
+| conforms to the GBNF grammar | **10 500 of 10 500 (100 %)** |
+| gate throughput | 4 200–4 470 documents/second, in-process |
+| size | 47 MiB train + 6.9 MiB holdout |
+
+Language elements across the corpus: **32 609** `пример`, **28 549** fields, **12 670**
+`правило`, **11 990** `объект`/`структура`, **7 166** `утилита`, **5 328** `морфизм`,
+**5 227** `свойство`, **1 638** `теорема`.
+
+### The three holdout strata
+
+The split divides the **space of templates and subject areas** and is written to
+`reports/split.json` *before the first document is generated*. A document is training data
+only if **both** its template and its area are on the training side.
+
+| stratum | documents | what is new |
+|---|---:|---|
+| `новая_область` | 501 | familiar structural form, unseen subject area |
+| `новый_шаблон` | 501 | unseen structural form, familiar area |
+| `новый_шаблон_новая_область` | 498 | **neither was ever seen** — the strict one |
+
+Measured leakage invariants (`reports/stats.json → leakage_check`, all hold): 0 shared
+documents, 0 shared task texts, 0 holdout templates leaked into train, 0 holdout areas
+leaked into train, and the strict stratum verifiably uses only unseen templates and only
+unseen areas.
+
+One constraint is enforced while building the split rather than checked afterwards: every
+tracked language construct must survive in at least one training template, or the model
+would never see it at all.
+
+## 3. The negative control — the part that is worth more than the corpus
+
+A dataset whose every row "passed the filter" tells you nothing until you know whether the
+filter can fail anything. So 4 400 known-good documents were corrupted on purpose, eleven
+ways, and put back through the same gate.
+
+**4 030 of 4 400 caught — 91.6 %.**
+
+| corruption | caught | what the gate said |
+|---|---:|---|
+| example: expected value changed | **100.0 %** | `FTS_EXAMPLE_MISMATCH` |
+| syntax: C-like (`field: type`) | **100.0 %** | `PARSE_ERROR` |
+| field not declared, used in a rule | **100.0 %** | `FTS_UTILITY_FIELD` |
+| example: required field dropped | **100.0 %** | `FTS_UTILITY_EXAMPLE_FIELD` |
+| indentation: four spaces instead of two | **100.0 %** | `PARSE_ERROR` |
+| theorem: unknown morphism | **100.0 %** | `PARSE_ERROR` |
+| type: string where a number belongs | **100.0 %** | `FTS_UTILITY_EXAMPLE_TYPE` |
+| utility: different return type | **100.0 %** | `FTS_UTILITY_INITIAL_TYPE` |
+| name: unclosed guillemet | 74.5 % | `PARSE_ERROR` |
+| **rule: threshold moved by one** | **72.8 %** | `FTS_EXAMPLE_MISMATCH` |
+| **morphism: reversed** | **60.2 %** | `PARSE_ERROR` |
+| **total** | **91.6 %** | |
+
+The last two rows are the limits of the method, and they are the reason this section
+exists:
+
+* **Threshold moved by one** is the only *semantic* corruption in the list — the document
+  stays well-formed and well-typed and simply computes something else. Before the
+  generator was changed, the gate caught **5.0 %** of these: the examples in a document
+  never happened to touch the boundary they were testing. Adding a pair of examples
+  "exactly at the threshold" and "one below it" moved it to **72.8 %**, and moved the
+  overall catch rate from **85.8 % to 91.6 %**. That change was made *because of this
+  measurement*, not before it — which is the argument for running a negative control at
+  all.
+* **A reversed morphism** is caught only when it breaks a theorem's type chain. In a
+  document with no theorem, a reversed morphism is still a perfectly correct document. The
+  reference toolchain records exactly this case in
+  `fts-gate/examples/fallacies/reversed-morphism.fts`.
+
+So the honest formulation, and the one the model card repeats:
+
+> The gate guarantees that a document compiles, type checks, executes its examples and
+> proves its theorem. It does **not** guarantee that the document means what was asked.
+
+### 3.1 The grammar is not the compiler
+
+The same 4 400 mutants were run through the GBNF recogniser: it **accepts 3 302 and
+rejects 1 098**. The grammar catches syntactic corruption and nothing else — type errors
+and semantic errors pass it cleanly. Constrained decoding is not a correctness check, and
+this is the measurement that says so.
+
+### 3.2 It ships as data, not as a table
+
+`negative_control/` contains all 4 400 mutated documents and the gate's verdict on each,
+so you can re-run the whole thing against your own checker instead of taking the table
+above on trust. It is deterministic (`random.Random(11)` over the shipped corpus) and was
+regenerated and re-verified for this publication: **4 030/4 400, every per-mutation count
+and every diagnostic code identical to the published summary.**
+
+## 4. Two other things that were measured about the gate
+
+**It rejected 17.2 % during development.** On the first full run the gate threw out
+**1 787 of 10 400** candidates, all `PARSE_ERROR`, all from a single bug: a name of the
+form `«Признак «есть медсправка» установлен»` terminated at the first closing guillemet.
+The compiler found it; review had not. The final run's 0 rejections are what a fixed
+generator looks like, not what a blind filter looks like.
+
+**The same gate on the pre-existing hand-written corpus.** Of 141 unique `.fts` documents
+found on the machine, **92 parse and pass**; 42 do not parse at all (module wrappers,
+fragments, deliberately broken examples from `fts-gate/examples/fallacies` and
+`refusals`), 6 fail on their examples, 1 on types. The gate is not tuned to accept
+everything it is shown.
+
+## 5. Coverage: this corpus is denser than the corpus it came from
+
+In those 92 hand-written documents, **11** of the 74 tracked constructs never appear at
+all and **21** appear in five documents or fewer. In this dataset the minimum is **127
+training documents** (`сравн:больше`) and **19 holdout documents**.
+
+| construct | hand-written corpus (92 docs) | train | holdout |
+|---|---:|---:|---:|
+| `вложена структура` | 0 | 243 | 21 |
+| `не равно` | 0 | 267 | 25 |
+| `затем применить морфизм` | 0 | 304 | 147 |
+| `получаем` | 0 | 392 | 153 |
+| `возвращает дату` | 0 | 396 | 21 |
+| `применить морфизм` | 0 | 589 | 166 |
+| `морфизм: из` / `в` | 0 | 786 | 302 |
+| `структура` | 0 | 882 | 61 |
+| `является текстом` | 0 | 1 131 | 80 |
+| `вложен объект` | 1 | 249 | 19 |
+| `является датой` | 2 | 1 035 | 169 |
+| `возвращает признак` | 3 | 1 182 | 61 |
+
+Full table for all 74 constructs: `reports/coverage_vs_corpus.json`.
+
+## 6. Row format
+
+```json
+{"id": "...", "split": "train", "stratum": null,
+ "prompt": "<task in Russian>",
+ "completion": "<the FTS specification>",
+ "meta": {"skeleton": "...", "domain": "...", "register": "...", "quote_style": "...",
+          "constructs": [...], "gate": {...}, "summary": {...}, "context": {...},
+          "sha256": "..."}}
+```
+
+The **system prompt is not duplicated into every row** — it is one line, it is identical
+for all 10 500, and it lives in `reports/stats.json → system_prompt`:
+
+```
+Ты пишешь исполняемые спецификации на языке FTS (русская поверхность). Ответ — только текст спецификации, без пояснений и без ограждений кода.
+```
+
+Build `messages` from that plus the `prompt`/`completion` pair.
+
+`meta.gate` records what the gate saw for that document, including
+`certificate_digest` for the 1 638 documents with a theorem. `meta.context` is the data a
+theorem is proved against, and is present only for those documents.
+
+### Task variety
+
+Task texts are written in **five registers**, roughly evenly distributed: `тз` (technical
+brief in prose), `тикет` (bulleted ticket), `письмо` (letter from a client), `регламент`
+(numbered clauses), `просьба` (a short imperative). **The registers are deliberately shared
+between train and holdout**: the holdout measures transfer to an unseen structural form and
+an unseen subject area, not to an unseen writing style.
+
+A task must fully determine its document — thresholds, initial value, properties and every
+test case are named explicitly — or the model would be learning to guess rather than to
+translate.
+
+## 7. Files and loading
+
+```
+data/train.jsonl                        9 000 rows, canonical
+data/holdout.jsonl                      1 500 rows, canonical, each with `stratum`
+hf/train.jsonl, hf/holdout.jsonl        the same rows, Arrow-safe (see below)
+hf/negative_control.jsonl               4 400 mutants joined to their verdicts
+negative_control/mutants.jsonl          4 400 corrupted documents
+negative_control/verdicts.jsonl         the gate's verdict on each
+negative_control/gate_sensitivity.json  the summary table of § 3
+reports/split.json                      the template/area split, written before generation
+reports/stats.json                      full statistics, coverage, leakage checks
+reports/coverage_vs_corpus.json         the language tail (§ 5)
+reports/grammar_conformance.json        10 500 / 10 500 accepted
+fts.gbnf                                the grammar, for constrained decoding
+MANIFEST.json                           sha256 of every published file
+```
+
+```python
+from datasets import load_dataset
+
+ds = load_dataset("digitable-lol/fts-specgen-dataset")                    # train, holdout
+nc = load_dataset("digitable-lol/fts-specgen-dataset", "negative_control")
+```
+
+**Why the dataset is here twice.** `meta` is a free-form object: `meta.context` alone takes
+**55 distinct key-sets** across the corpus — one per subject area — and `meta` itself has
+two key-sets, because only documents with a theorem carry `context`.
+
+On **`datasets` ≥ 5 this is fine**, and that was checked rather than assumed:
+`load_dataset("json", data_files="data/train.jsonl")` was run on `datasets` 5.0.1, loads
+all 9 000 rows, and types the column as `Json`, handing back a plain dict. Earlier versions
+infer an Arrow struct from a sample of rows and fail on the union.
+
+`hf/` sidesteps the question entirely and is what the dataset viewer and the `configs`
+above point at: `meta_json` carries the original object as a string (`json.loads`
+round-trips it exactly), alongside lifted scalar columns `skeleton`, `domain`, `register`,
+`quote_style`, `n_utilities`, `n_examples`, `n_theorems`, `has_context`, so you can filter
+without parsing. **Train on `data/`; browse and filter with `hf/`.** The same problem and
+the same fix are documented in
+[`digitable-lol/digit-router-dataset`](https://huggingface.co/datasets/digitable-lol/digit-router-dataset).
+
+## 8. Reproduction
+
+```bash
+python3 tools/generate.py --train 9000 --holdout 1500
+node    tools/verify.mjs < dataset/candidates.jsonl > work/verdicts.jsonl
+python3 tools/assemble.py
+python3 tools/check_dataset.py                 # independent re-check: 10500/10500
+node    tools/grammar_check.mjs dataset/train.jsonl dataset/holdout.jsonl
+
+# the negative control
+python3 tools/mutate.py work/final.jsonl work/mutants.jsonl 400
+node    tools/verify.mjs < work/mutants.jsonl > work/mutant_verdicts.jsonl
+```
+
+The split is deterministic (`SPLIT_SEED = 20260806`) and so is generation
+(`random.Random(f"{seed}:{skeleton}:{domain}:{round}")`), so a re-run reproduces the same
+files. Reproducing the gate verdicts requires the FTS compiler.
+
+## 9. Limits — read before using
+
+1. **Semantic fidelity is unmeasured** (§ 1.1). The largest gap.
+2. **No document combines a utility with a theorem** (§ 1.2). A measured hole with a
+   measured downstream consequence.
+3. **Synthetic throughout.** Task texts are composed from registers and templates, not
+   collected from people. Indirect, elliptical and under-specified requests — the normal
+   way humans ask for things — are absent by construction, because a task here must fully
+   determine its document.
+4. **Russian only**, and one language's Russian surface (§ 1.3).
+5. **33 templates and 62 areas** is not the space of specifications anyone might want. It
+   is the space this generator covers.
+6. **The gate's own blind spots are 8.4 %**, itemised in § 3, and they are concentrated in
+   exactly the corruption that matters most: a semantically wrong threshold in a
+   syntactically perfect document.
+7. **Coverage warnings are not rejections.** `FTS_COVERAGE_HOLE`, `FTS_RULE_OVERLAP`,
+   `FTS_PROPERTY_UNATTAINABLE` and `FTS_PROPERTY_VIOLATED` are recorded in
+   `meta.gate.coverage_warnings` but do not fail a document. This is measured, not assumed:
+   the toolchain's own reference example
+   `flang-v6/examples/utilities/discount.fts` raises `FTS_PROPERTY_VIOLATED` and
+   `FTS_PROPERTY_UNATTAINABLE` itself, because the language has no way to declare that a
+   money field is non-negative, so the analyser always finds a negative input. Rejecting on
+   that signal would mean rejecting an idiom of the language.
+
+## 10. Origin and licensing
+
+**BSD-2-Clause.**
+
+The FTS language and its compiler
+([`@digitable-lol/fts`](https://github.com/digitable-lol/flang)) and the verification gate
+(`@digitable-lol/fts-gate`) are both BSD-2-Clause and both belong to this organisation.
+The corpus was generated from them by this project's own code, and `fts.gbnf` is shipped
+here under the same terms.
+
+No teacher model was used anywhere in the pipeline. Every document was constructed
+programmatically and then checked by the compiler; expected example values come from the
+language's own reference interpreter and property bounds from conservative analytic
+estimates.
+
+There is no third-party copyleft question here, which is a deliberate contrast with
+[`digitable-lol/digit-router-dataset`](https://huggingface.co/datasets/digitable-lol/digit-router-dataset),
+where there is one and where this organisation says so rather than picking a convenient tag.
+
+## 11. Integrity
+
+`MANIFEST.json` lists the sha256 of every published file, recorded at upload time on the
+machine that produced it. Each document additionally carries its own `meta.sha256`.
+
+---
+
+## 12. По-русски
+
+**Что это.** 10 500 пар «задание на обычном русском → исполняемая спецификация FTS»:
+9 000 train, 1 500 holdout. **Каждый документ прогнан через настоящий компилятор FTS** —
+разобран, типизирован, его примеры исполнены, его теорема доказана и сертифицирована; в
+файлы попало только то, что прошло гейт целиком. Это те данные, на которых обучен
+[`digitable-lol/specgen-qwen3-1.7b`](https://huggingface.co/digitable-lol/specgen-qwen3-1.7b).
+
+**Главная оговорка.** «Прошло компилятор» — утверждение о форме, не о смысле. Гейт
+доказывает, что документ компилируется, типизируется, исполняет свои примеры и доказывает
+свою теорему. Что он означает ровно то, что просили словами в задании, — **не измерено**.
+Генератор строит задание и документ из одной структуры, поэтому расхождение возможно
+только как ошибка в `tools/task.py`, и её масштаб неизвестен.
+
+**Дыра в датасете, найденная обученной на нём моделью.** **Ни один из 9 000 обучающих
+документов не содержит утилиту и теорему одновременно**: 6 451 с утилитой без теоремы,
+1 185 с теоремой без утилиты, 1 364 без того и другого, **ноль** с обоими. В holdout
+комбинация есть — шаблон `CMB-UTL-THM`, 129 документов. Последствие проявилось в модели:
+из 127 таких документов, которые она написала и которые чисто скомпилировались, теорему
+содержат **0**, при 324/324 на остальных теоремах holdout. Компилятор здесь бессилен:
+документ без теоремы формально валиден. Это дефект данных, и он тут написан.
+
+**Негативный контроль — самое ценное в этой публикации.** Фраза «все примеры прошли
+фильтр» ничего не значит, пока не известно, способен ли фильтр вообще что-то отвергнуть.
+4 400 заведомо исправных документов испорчены намеренно — **одиннадцатью способами** — и
+пропущены через тот же гейт. **Поймано 4 030 из 4 400 — 91,6 %.** Восемь видов порчи
+ловятся на 100 %; три — нет: незакрытая ёлочка 74,5 %, **сдвинутый на единицу порог
+72,8 %**, **перевёрнутый морфизм 60,2 %**.
+
+Две последние строки — граница метода. *Сдвинутый порог* — единственная чисто смысловая
+порча в списке: документ остаётся корректным и просто считает другое. До появления
+граничных примеров ловилось **5,0 %** — примеры просто не задевали проверяемую границу;
+после добавления пары «ровно на пороге» и «на единицу ниже» стало 72,8 %, а общая доля
+выросла с **85,8 % до 91,6 %**. Это изменение сделано **по результату замера**, а не
+заранее, и в этом весь смысл негативного контроля. *Перевёрнутый морфизм* ловится только
+когда ломает типовую цепочку теоремы: в документе без теоремы он остаётся корректным
+документом.
+
+**Негативный контроль опубликован данными, а не таблицей:** все 4 400 испорченных
+документов и вердикт гейта по каждому лежат в `negative_control/`, так что таблицу выше
+можно перепроверить своим инструментом. Он детерминирован и был перегенерирован и
+перепроверен при подготовке этой публикации — **4 030/4 400, все счётчики и все коды
+диагностик совпали с опубликованной сводкой**.
+
+**Грамматика — не компилятор.** Тот же распознаватель GBNF на тех же 4 400 мутантах
+принимает 3 302 и отвергает 1 098: синтаксическую порчу он ловит, типовую и смысловую —
+нет. Constrained decoding не является проверкой корректности, и это то измерение, которое
+об этом говорит.
+
+**Разделение — до генерации.** `reports/split.json` делит пространство шаблонов и
+областей и пишется **до построения первого документа**. Три страты: `новая_область` (501),
+`новый_шаблон` (501), `новый_шаблон_новая_область` (498, самая строгая). Утечка измерена и
+равна нулю по всем пяти инвариантам.
+
+**Что ещё измерено про гейт.** По ходу разработки он отверг **1 787 из 10 400 (17,2 %)** —
+все `PARSE_ERROR` из-за одной ошибки в именах с вложенными ёлочками; нашёл её компилятор,
+а не ревью. Тот же гейт на исходном рукописном корпусе: из 141 документа проходят **92**,
+42 не разбираются вовсе, 6 падают на примерах, 1 — на типах. Гейт не устроен так, чтобы
+принимать всё подряд.
+
+**Ограничения:** смысловая верность не измерена; нет ни одного документа с утилитой и
+теоремой сразу; всё синтетическое — живых пользовательских формулир��вок нет по построению;
+только русский; 33 шаблона и 62 области — это охват генератора, а не пространство
+спецификаций; слепые зоны гейта — 8,4 %, и они сосредоточены ровно там, где важнее всего.
+
+**Лицензия — BSD-2-Clause.** Язык FTS, компилятор и гейт принадлежат этой же организации и
+лицензированы так же; teacher-модель не использовалась нигде. Нерешённого вопроса о
+копилефте здесь нет — в отличие от
+[`digit-router-dataset`](https://huggingface.co/datasets/digitable-lol/digit-router-dataset),
+где он есть и где мы это пишем прямо, а не выбираем удобный тег.

data/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d50c09795840a8548010f3ecf7ccc460b3c1be60ee8ba4c071a775ff0abdacc2
+size 48929988

fts.gbnf

@@ -0,0 +1,151 @@
+# FTS natural (Russian) surface — GBNF grammar for constrained decoding.
+#
+# Derived from the natural-surface parser of @digitable/fts. Scope and known gaps are
+# documented in grammars/README.md; verify with `npm run grammar:check`.
+#
+# Каждое правило — в одну строку, какой бы длинной она ни вышла. Перенос
+# альтернативы на следующую строку читается лучше, но llama.cpp завершает
+# правило на переводе строки вне скобок и отказывается грузить файл целиком:
+# «expecting name at |». Наш чекер этого не ловил — он разбирает грамматику
+# своим парсером и на переносы не смотрит, поэтому зелёный `grammar:check`
+# ничего про пригодность к constrained decoding не говорил. Замер: до склейки
+# двух таких строк llama.cpp не грузил грамматику вовсе; после — грузит, и на
+# одной и той же модели синтаксически чистых генераций стало 25/25 вместо
+# 17/25. Проверку на продолжения добавили в scripts/check-grammar.mjs.
+#
+# Covered: категория / объект / структура / морфизм / утилита / теорема,
+# including правило, свойство, пример, all comparison phrases, percent and
+# field operands, guillemet and bare names, `//` comments and blank lines.
+#
+# Deliberate restrictions relative to the parser (a strict subset — every
+# string this grammar accepts is accepted by the parser, not the converse):
+#   * indentation is exactly two spaces per level, never tabs;
+#   * the English surface and the legacy braced syntax are out of scope;
+#   * `принимает` / `возвращает` / `начинает с` appear in that fixed order;
+#   * at most one теорема, and it comes last;
+#   * `/* … */` block comments are out of scope.
+
+root ::= skip document
+
+document ::= category-line body
+
+category-line ::= "категория " name eol
+
+body ::= declaration* theorem?
+
+declaration ::= object | morphism | utility
+
+# ---------------------------------------------------------------- объекты
+
+object ::= ("  объект " | "  структура ") name eol field+
+
+field ::= plain-field | nested-field
+
+plain-field ::= "    " name (" является " | " иногда является ") field-type eol
+
+# A field type is a built-in, a state witness type, or any nominal type name.
+# The parser accepts arbitrary nominal types (`является Email`), so the grammar
+# does too: it constrains syntax, never the domain vocabulary.
+field-type ::= builtin-type | state-type | name
+
+builtin-type ::= "строкой" | "текстом" | "числом" | "датой" | "деньгами" | "признаком"
+
+state-type ::= "состоянием " name
+
+nested-field ::= "    " ("вложен объект " | "вложена структура ") name eol
+
+# --------------------------------------------------------------- морфизмы
+
+morphism ::= "  морфизм " name eol morphism-domain morphism-codomain morphism-law?
+
+morphism-domain ::= ("    если " | "    из ") name eol
+
+morphism-codomain ::= ("    то " | "    в ") name eol
+
+morphism-law ::= "    по закону " name eol
+
+# --------------------------------------------------------------- утилиты
+
+utility ::= "  утилита " name eol utility-accepts utility-returns utility-initial utility-member+
+
+utility-accepts ::= "    принимает " name eol
+
+utility-returns ::= "    возвращает " return-type eol
+
+return-type ::= "строку" | "число" | "дату" | "деньги" | "признак"
+
+utility-initial ::= "    начинает с " scalar eol
+
+utility-member ::= rule | property | example
+
+rule ::= "    правило " name eol rule-if rule-and* rule-then
+
+rule-if ::= "      если " name " " comparison eol
+
+rule-and ::= "      и " name " " comparison eol
+
+rule-then ::= rule-add | rule-set
+
+rule-add ::= "      то добавить " operand eol
+
+rule-set ::= "      то результат " equality " " operand eol
+
+property ::= "    свойство " name eol "      результат " comparison eol
+
+example ::= "    пример " name eol given+ expected
+
+given ::= "      дано " name " " equality " " scalar eol
+
+expected ::= "      ожидается результат " equality " " scalar eol
+
+comparison ::= comparison-operator " " operand
+
+comparison-operator ::= "не меньше" | "не больше" | "не равен" | "не равна" | "не равно" | "равен" | "равна" | "равно" | "больше" | "меньше"
+
+equality ::= "равен" | "равна" | "равно"
+
+operand ::= result-operand | percent-operand | field-operand | scalar
+
+result-operand ::= "результат"
+
+percent-operand ::= number " " ("процентов" | "процента" | "процент") " от поля " name
+
+field-operand ::= "поле " name
+
+# --------------------------------------------------------------- теоремы
+
+theorem ::= "  теорема " name eol theorem-given theorem-lookup theorem-step* theorem-conclusion?
+
+theorem-given ::= "    дано " name " имеет " name " равное " scalar eol
+
+theorem-lookup ::= "    в данных " name " найти где " name " равен " scalar eol
+
+theorem-step ::= ("    по морфизму " | "    затем по морфизму " | "    применить морфизм " | "    затем применить морфизм ") name eol
+
+theorem-conclusion ::= ("    следовательно " | "    получаем ") name eol
+
+# ------------------------------------------------------- имена и значения
+
+# A name is either a guillemet-quoted phrase (any characters but the closing
+# guillemet and a newline) or a single bare identifier without spaces.
+name ::= quoted-name | bare-name
+
+quoted-name ::= "«" [^«»\n]+ "»"
+
+bare-name ::= [a-zA-Zа-яёА-ЯЁ_] [a-zA-Z0-9а-яёА-ЯЁ_-]*
+
+scalar ::= boolean | null-value | number | name
+
+boolean ::= "да" | "нет"
+
+null-value ::= "ничто"
+
+number ::= "-"? ("0" | [1-9] [0-9]*) ("." [0-9]+)?
+
+# -------------------------------------------------------- строки и пробелы
+
+# Every logical line ends here. Blank lines and `//` comment lines are
+# discarded by the parser, so the grammar lets them appear between any lines.
+eol ::= "\n" skip
+
+skip ::= (" "* "//" [^\n]* "\n" | "\n")*

hf/negative_control.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9808a3afa16f2ed6beb67479f1d5f2d4a4843dd22a11f73fc3fda0b0f61e7ee
+size 12016316

hf/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f645f250af1eca9cc9a3b01065dc4deb287ad5dcbc39a73a395c05c8f703910
+size 52279316

negative_control/gate_sensitivity.json

@@ -0,0 +1,95 @@
+{
+  "total": 4400,
+  "caught": 4030,
+  "catch_rate": 0.9159,
+  "by_mutation": {
+    "пример: другое ожидание": {
+      "total": 400,
+      "caught": 400,
+      "rate": 1.0,
+      "codes": {
+        "examples:FTS_EXAMPLE_MISMATCH": 400
+      }
+    },
+    "синтаксис: C-подобный": {
+      "total": 400,
+      "caught": 400,
+      "rate": 1.0,
+      "codes": {
+        "compile:PARSE_ERROR": 400
+      }
+    },
+    "поле: незадекларированное в правиле": {
+      "total": 400,
+      "caught": 400,
+      "rate": 1.0,
+      "codes": {
+        "validate:FTS_UTILITY_FIELD": 400
+      }
+    },
+    "морфизм: перевёрнут": {
+      "total": 400,
+      "caught": 241,
+      "rate": 0.6025,
+      "codes": {
+        "compile:PARSE_ERROR": 241
+      }
+    },
+    "пример: пропущено обязательное поле": {
+      "total": 400,
+      "caught": 400,
+      "rate": 1.0,
+      "codes": {
+        "validate:FTS_UTILITY_EXAMPLE_FIELD": 400
+      }
+    },
+    "отступы: четыре пробела": {
+      "total": 400,
+      "caught": 400,
+      "rate": 1.0,
+      "codes": {
+        "compile:PARSE_ERROR": 400
+      }
+    },
+    "имя: незакрытая ёлочка": {
+      "total": 400,
+      "caught": 298,
+      "rate": 0.745,
+      "codes": {
+        "compile:PARSE_ERROR": 298
+      }
+    },
+    "теорема: неизвестный морфизм": {
+      "total": 400,
+      "caught": 400,
+      "rate": 1.0,
+      "codes": {
+        "compile:PARSE_ERROR": 400
+      }
+    },
+    "тип: строка вместо числа в примере": {
+      "total": 400,
+      "caught": 400,
+      "rate": 1.0,
+      "codes": {
+        "validate:FTS_UTILITY_EXAMPLE_TYPE": 400
+      }
+    },
+    "утилита: другой возвращаемый тип": {
+      "total": 400,
+      "caught": 400,
+      "rate": 1.0,
+      "codes": {
+        "validate:FTS_UTILITY_INITIAL_TYPE": 400
+      }
+    },
+    "правило: порог сдвинут (семантика, не синтаксис)": {
+      "total": 400,
+      "caught": 291,
+      "rate": 0.7275,
+      "codes": {
+        "examples:FTS_EXAMPLE_MISMATCH": 291
+      }
+    }
+  }
+}

negative_control/mutants.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af893c1c07d43e1c91f9b61b2a350ce1f347d03d6ca02e77de12bf94c4334e4c
+size 11066486

reports/coverage_vs_corpus.json

@@ -0,0 +1,362 @@
+{
+  "категория": {
+    "corpus_docs": 77,
+    "train": 9000,
+    "holdout": 1500
+  },
+  "объект": {
+    "corpus_docs": 77,
+    "train": 8118,
+    "holdout": 1439
+  },
+  "структура": {
+    "corpus_docs": 0,
+    "train": 882,
+    "holdout": 61
+  },
+  "морфизм": {
+    "corpus_docs": 31,
+    "train": 1972,
+    "holdout": 624
+  },
+  "утилита": {
+    "corpus_docs": 44,
+    "train": 6451,
+    "holdout": 715
+  },
+  "теорема": {
+    "corpus_docs": 21,
+    "train": 1185,
+    "holdout": 453
+  },
+  "является": {
+    "corpus_docs": 77,
+    "train": 9000,
+    "holdout": 1500
+  },
+  "иногда является": {
+    "corpus_docs": 2,
+    "train": 397,
+    "holdout": 127
+  },
+  "вложен объект": {
+    "corpus_docs": 1,
+    "train": 249,
+    "holdout": 19
+  },
+  "вложена структура": {
+    "corpus_docs": 0,
+    "train": 243,
+    "holdout": 21
+  },
+  "тип:строкой": {
+    "corpus_docs": 45,
+    "train": 3098,
+    "holdout": 950
+  },
+  "тип:текстом": {
+    "corpus_docs": 0,
+    "train": 1131,
+    "holdout": 80
+  },
+  "тип:числом": {
+    "corpus_docs": 25,
+    "train": 6653,
+    "holdout": 730
+  },
+  "тип:датой": {
+    "corpus_docs": 2,
+    "train": 1035,
+    "holdout": 169
+  },
+  "тип:деньгами": {
+    "corpus_docs": 31,
+    "train": 4568,
+    "holdout": 483
+  },
+  "тип:признаком": {
+    "corpus_docs": 32,
+    "train": 3532,
+    "holdout": 421
+  },
+  "тип:состоянием": {
+    "corpus_docs": 34,
+    "train": 1972,
+    "holdout": 744
+  },
+  "тип:номинальный": {
+    "corpus_docs": 3,
+    "train": 633,
+    "holdout": 44
+  },
+  "морфизм:если": {
+    "corpus_docs": 31,
+    "train": 1186,
+    "holdout": 322
+  },
+  "морфизм:то": {
+    "corpus_docs": 31,
+    "train": 1186,
+    "holdout": 322
+  },
+  "морфизм:из": {
+    "corpus_docs": 0,
+    "train": 786,
+    "holdout": 302
+  },
+  "морфизм:в": {
+    "corpus_docs": 0,
+    "train": 786,
+    "holdout": 302
+  },
+  "по закону": {
+    "corpus_docs": 25,
+    "train": 396,
+    "holdout": 539
+  },
+  "принимает": {
+    "corpus_docs": 44,
+    "train": 6451,
+    "holdout": 715
+  },
+  "возвращает:строку": {
+    "corpus_docs": 1,
+    "train": 393,
+    "holdout": 150
+  },
+  "возвращает:число": {
+    "corpus_docs": 17,
+    "train": 1584,
+    "holdout": 209
+  },
+  "возвращает:дату": {
+    "corpus_docs": 0,
+    "train": 396,
+    "holdout": 21
+  },
+  "возвращает:деньги": {
+    "corpus_docs": 25,
+    "train": 2896,
+    "holdout": 274
+  },
+  "возвращает:признак": {
+    "corpus_docs": 3,
+    "train": 1182,
+    "holdout": 61
+  },
+  "начинает с": {
+    "corpus_docs": 44,
+    "train": 6451,
+    "holdout": 715
+  },
+  "правило": {
+    "corpus_docs": 44,
+    "train": 6451,
+    "holdout": 715
+  },
+  "свойство": {
+    "corpus_docs": 33,
+    "train": 3690,
+    "holdout": 569
+  },
+  "пример": {
+    "corpus_docs": 42,
+    "train": 6451,
+    "holdout": 715
+  },
+  "правило:если": {
+    "corpus_docs": 44,
+    "train": 6451,
+    "holdout": 715
+  },
+  "правило:и": {
+    "corpus_docs": 19,
+    "train": 397,
+    "holdout": 20
+  },
+  "то добавить": {
+    "corpus_docs": 34,
+    "train": 4480,
+    "holdout": 483
+  },
+  "то результат": {
+    "corpus_docs": 17,
+    "train": 1971,
+    "holdout": 232
+  },
+  "свойство:результат": {
+    "corpus_docs": 33,
+    "train": 3690,
+    "holdout": 569
+  },
+  "дано(пример)": {
+    "corpus_docs": 42,
+    "train": 6451,
+    "holdout": 715
+  },
+  "ожидается результат": {
+    "corpus_docs": 42,
+    "train": 6451,
+    "holdout": 715
+  },
+  "сравн:не меньше": {
+    "corpus_docs": 35,
+    "train": 5665,
+    "holdout": 547
+  },
+  "сравн:не больше": {
+    "corpus_docs": 32,
+    "train": 2108,
+    "holdout": 233
+  },
+  "сравн:не равен": {
+    "corpus_docs": 6,
+    "train": 591,
+    "holdout": 159
+  },
+  "сравн:не равна": {
+    "corpus_docs": 0,
+    "train": 323,
+    "holdout": 45
+  },
+  "сравн:не равно": {
+    "corpus_docs": 0,
+    "train": 267,
+    "holdout": 25
+  },
+  "сравн:равен": {
+    "corpus_docs": 62,
+    "train": 7636,
+    "holdout": 1039
+  },
+  "сравн:равна": {
+    "corpus_docs": 25,
+    "train": 4672,
+    "holdout": 433
+  },
+  "сравн:равно": {
+    "corpus_docs": 18,
+    "train": 2703,
+    "holdout": 370
+  },
+  "сравн:больше": {
+    "corpus_docs": 24,
+    "train": 127,
+    "holdout": 134
+  },
+  "сравн:меньше": {
+    "corpus_docs": 12,
+    "train": 1190,
+    "holdout": 59
+  },
+  "операнд:результат": {
+    "corpus_docs": 2,
+    "train": 393,
+    "holdout": 21
+  },
+  "операнд:процентов от поля": {
+    "corpus_docs": 25,
+    "train": 1650,
+    "holdout": 204
+  },
+  "операнд:процента от поля": {
+    "corpus_docs": 6,
+    "train": 1316,
+    "holdout": 112
+  },
+  "операнд:процент от поля": {
+    "corpus_docs": 1,
+    "train": 395,
+    "holdout": 20
+  },
+  "операнд:поле": {
+    "corpus_docs": 9,
+    "train": 524,
+    "holdout": 30
+  },
+  "теорема:дано": {
+    "corpus_docs": 21,
+    "train": 1185,
+    "holdout": 453
+  },
+  "в данных": {
+    "corpus_docs": 21,
+    "train": 1185,
+    "holdout": 453
+  },
+  "по морфизму": {
+    "corpus_docs": 21,
+    "train": 596,
+    "holdout": 287
+  },
+  "затем по морфизму": {
+    "corpus_docs": 6,
+    "train": 697,
+    "holdout": 296
+  },
+  "применить морфизм": {
+    "corpus_docs": 0,
+    "train": 589,
+    "holdout": 166
+  },
+  "затем применить морфизм": {
+    "corpus_docs": 0,
+    "train": 304,
+    "holdout": 147
+  },
+  "следовательно": {
+    "corpus_docs": 21,
+    "train": 793,
+    "holdout": 175
+  },
+  "получаем": {
+    "corpus_docs": 0,
+    "train": 392,
+    "holdout": 153
+  },
+  "скаляр:да": {
+    "corpus_docs": 44,
+    "train": 4781,
+    "holdout": 637
+  },
+  "скаляр:нет": {
+    "corpus_docs": 26,
+    "train": 3569,
+    "holdout": 309
+  },
+  "скаляр:ничто": {
+    "corpus_docs": 3,
+    "train": 390,
+    "holdout": 23
+  },
+  "скаляр:число": {
+    "corpus_docs": 45,
+    "train": 6451,
+    "holdout": 715
+  },
+  "скаляр:отрицательное число": {
+    "corpus_docs": 8,
+    "train": 394,
+    "holdout": 20
+  },
+  "скаляр:дробное число": {
+    "corpus_docs": 4,
+    "train": 1305,
+    "holdout": 325
+  },
+  "имя:«ёлочки»": {
+    "corpus_docs": 78,
+    "train": 8996,
+    "holdout": 1499
+  },
+  "имя:голое": {
+    "corpus_docs": 54,
+    "train": 2369,
+    "holdout": 380
+  },
+  "комментарий //": {
+    "corpus_docs": 76,
+    "train": 5058,
+    "holdout": 828
+  }
+}

reports/grammar_conformance.json

@@ -0,0 +1,8 @@
+{
+  "total": 10500,
+  "accepted": 10500,
+  "rejected": 0,
+  "samples": [],
+  "accept_rate": 1,
+  "rejected_by_skeleton": {}
+}

reports/split.json

@@ -0,0 +1,162 @@
+{
+  "seed": 20260806,
+  "rule": "train = train-скелет И train-область; всё прочее — holdout",
+  "holdout_skeletons": [
+    "CMB-UTL-THM",
+    "MOR-FROM-TO-LAW",
+    "OBJ-OPTIONAL",
+    "OBJ-STATES",
+    "THM-APPLY",
+    "THM-NOCONCL",
+    "UTL-NUM-FRACT",
+    "UTL-STR-BYTEXT"
+  ],
+  "train_skeletons": [
+    "CMB-MOR-UTL",
+    "CMB-STRUCT-UTL",
+    "CMB-TWO-UTIL",
+    "MOR-IF-THEN",
+    "OBJ-BUILTIN",
+    "OBJ-NESTED",
+    "OBJ-NESTED-STRUCT",
+    "OBJ-TEXT-DATE",
+    "THM-CLASSIC",
+    "THM-LONG",
+    "THM-SHORT",
+    "UTL-DATE",
+    "UTL-FLAG-NEGATIVE",
+    "UTL-FLAG-NOTNULL",
+    "UTL-FLAG-SET",
+    "UTL-MONEY-FIELD",
+    "UTL-MONEY-FRACT",
+    "UTL-MONEY-MULTICOND",
+    "UTL-MONEY-NEG",
+    "UTL-MONEY-PCT",
+    "UTL-MONEY-PCT-SINGULAR",
+    "UTL-NUM-CONST",
+    "UTL-NUM-FIELDPROP",
+    "UTL-NUM-RESULT",
+    "UTL-STR-CLASSIFY"
+  ],
+  "holdout_domains": [
+    "agro",
+    "datacenter",
+    "gov",
+    "gym",
+    "insurance",
+    "livestock",
+    "museum",
+    "pharmacy",
+    "print",
+    "retail",
+    "telecom",
+    "theatre"
+  ],
+  "train_domains": [
+    "ads",
+    "airline",
+    "archive",
+    "autoservice",
+    "banking_ops",
+    "carsharing",
+    "cinema",
+    "clinic",
+    "construction",
+    "credit",
+    "customs",
+    "delivery_food",
+    "ecom",
+    "energy",
+    "forestry",
+    "fuel",
+    "hiring",
+    "hosting",
+    "hotel",
+    "hr",
+    "infosec",
+    "isp",
+    "itsm",
+    "lab",
+    "leasing",
+    "legal",
+    "library",
+    "logistics",
+    "manufacturing",
+    "mfc",
+    "notary",
+    "parking",
+    "payroll",
+    "procurement",
+    "qc",
+    "rail",
+    "realty",
+    "restaurant",
+    "saas",
+    "school",
+    "security",
+    "sport",
+    "tax",
+    "taxi",
+    "tourism",
+    "university",
+    "utilities_jkh",
+    "warehouse",
+    "waste",
+    "water"
+  ],
+  "constructs_covered_by_train_skeletons": [
+    "в данных",
+    "вложен объект",
+    "вложена структура",
+    "возвращает:дату",
+    "возвращает:деньги",
+    "возвращает:признак",
+    "возвращает:строку",
+    "возвращает:число",
+    "затем по морфизму",
+    "затем применить морфизм",
+    "иногда является",
+    "морфизм",
+    "морфизм:в",
+    "морфизм:если",
+    "морфизм:из",
+    "морфизм:то",
+    "объект",
+    "операнд:поле",
+    "операнд:процент от поля",
+    "операнд:процента от поля",
+    "операнд:процентов от поля",
+    "операнд:результат",
+    "по закону",
+    "по морфизму",
+    "получаем",
+    "правило:и",
+    "применить морфизм",
+    "пример",
+    "свойство",
+    "скаляр:да",
+    "скаляр:дробное число",
+    "скаляр:нет",
+    "скаляр:ничто",
+    "скаляр:отрицательное число",
+    "следовательно",
+    "сравн:больше",
+    "сравн:меньше",
+    "сравн:не больше",
+    "сравн:не равен",
+    "структура",
+    "теорема",
+    "теорема:дано",
+    "тип:датой",
+    "тип:деньгами",
+    "тип:номинальный",
+    "тип:признаком",
+    "тип:состоянием",
+    "тип:строкой",
+    "тип:текстом",
+    "тип:числом",
+    "то добавить",
+    "то результат",
+    "утилита"
+  ]
+}

reports/stats.json

@@ -0,0 +1,412 @@
+{
+  "system_prompt": "Ты пишешь исполняемые спецификации на языке FTS (русская поверхность). Ответ — только текст спецификации, без пояснений и без ограждений кода.",
+  "generated_candidates": 10500,
+  "passed_gate": 10500,
+  "rejected_by_gate": 0,
+  "gate_pass_rate": 1.0,
+  "train": 9000,
+  "holdout": 1500,
+  "gate_verdicts": {
+    "ok:": 10500
+  },
+  "leakage_check": {
+    "shared_document_hashes": 0,
+    "shared_prompt_hashes": 0,
+    "holdout_skeletons_leaked_into_train": [],
+    "holdout_domains_leaked_into_train": [],
+    "strict_stratum_size": 498,
+    "strict_stratum_uses_only_unseen_skeletons": true,
+    "strict_stratum_uses_only_unseen_domains": true
+  },
+  "by_stratum": {
+    "новая_область": 501,
+    "новый_шаблон": 501,
+    "новый_шаблон_новая_область": 498
+  },
+  "by_skeleton": {
+    "train": {
+      "CMB-TWO-UTIL": 396,
+      "UTL-MONEY-MULTICOND": 397,
+      "UTL-MONEY-FRACT": 394,
+      "UTL-MONEY-PCT": 399,
+      "OBJ-BUILTIN": 237,
+      "UTL-FLAG-SET": 396,
+      "UTL-DATE": 396,
+      "OBJ-TEXT-DATE": 242,
+      "THM-LONG": 396,
+      "MOR-IF-THEN": 393,
+      "UTL-NUM-CONST": 397,
+      "UTL-MONEY-FIELD": 127,
+      "CMB-MOR-UTL": 394,
+      "OBJ-NESTED": 249,
+      "UTL-STR-CLASSIFY": 393,
+      "UTL-MONEY-PCT-SINGULAR": 395,
+      "UTL-FLAG-NOTNULL": 390,
+      "UTL-FLAG-NEGATIVE": 396,
+      "OBJ-NESTED-STRUCT": 243,
+      "THM-SHORT": 392,
+      "UTL-NUM-RESULT": 393,
+      "UTL-NUM-FIELDPROP": 397,
+      "UTL-MONEY-NEG": 394,
+      "CMB-STRUCT-UTL": 397,
+      "THM-CLASSIC": 397
+    },
+    "holdout": {
+      "CMB-STRUCT-UTL": 19,
+      "OBJ-STATES": 120,
+      "THM-LONG": 23,
+      "OBJ-NESTED-STRUCT": 21,
+      "THM-NOCONCL": 125,
+      "MOR-FROM-TO-LAW": 130,
+      "OBJ-OPTIONAL": 108,
+      "UTL-MONEY-NEG": 20,
+      "UTL-STR-BYTEXT": 127,
+      "CMB-UTL-THM": 129,
+      "UTL-DATE": 21,
+      "UTL-NUM-FRACT": 128,
+      "OBJ-BUILTIN": 20,
+      "UTL-MONEY-PCT-SINGULAR": 20,
+      "UTL-NUM-FIELDPROP": 20,
+      "OBJ-NESTED": 19,
+      "OBJ-TEXT-DATE": 21,
+      "UTL-MONEY-MULTICOND": 20,
+      "THM-APPLY": 132,
+      "CMB-TWO-UTIL": 20,
+      "UTL-FLAG-NEGATIVE": 18,
+      "MOR-IF-THEN": 22,
+      "UTL-MONEY-FIELD": 6,
+      "UTL-NUM-RESULT": 21,
+      "THM-SHORT": 21,
+      "UTL-NUM-CONST": 21,
+      "UTL-FLAG-NOTNULL": 23,
+      "UTL-STR-CLASSIFY": 23,
+      "THM-CLASSIC": 23,
+      "UTL-FLAG-SET": 20,
+      "UTL-MONEY-FRACT": 22,
+      "CMB-MOR-UTL": 19,
+      "UTL-MONEY-PCT": 18
+    }
+  },
+  "by_domain_count": {
+    "train": 50,
+    "holdout": 62
+  },
+  "by_register": {
+    "train": {
+      "просьба": 1882,
+      "тз": 1876,
+      "регламент": 1748,
+      "тикет": 1787,
+      "письмо": 1707
+    },
+    "holdout": {
+      "тикет": 255,
+      "письмо": 226,
+      "просьба": 367,
+      "тз": 326,
+      "регламент": 326
+    }
+  },
+  "language_elements_total": {
+    "structures": 11990,
+    "fields": 28549,
+    "functors": 5328,
+    "utilities": 7166,
+    "rules": 12670,
+    "properties": 5227,
+    "examples": 32609,
+    "theorem": 1638
+  },
+  "construct_coverage": {
+    "категория": {
+      "train": 9000,
+      "holdout": 1500
+    },
+    "объект": {
+      "train": 8118,
+      "holdout": 1439
+    },
+    "структура": {
+      "train": 882,
+      "holdout": 61
+    },
+    "морфизм": {
+      "train": 1972,
+      "holdout": 624
+    },
+    "утилита": {
+      "train": 6451,
+      "holdout": 715
+    },
+    "теорема": {
+      "train": 1185,
+      "holdout": 453
+    },
+    "является": {
+      "train": 9000,
+      "holdout": 1500
+    },
+    "иногда является": {
+      "train": 397,
+      "holdout": 127
+    },
+    "вложен объект": {
+      "train": 249,
+      "holdout": 19
+    },
+    "вложена структура": {
+      "train": 243,
+      "holdout": 21
+    },
+    "тип:строкой": {
+      "train": 3098,
+      "holdout": 950
+    },
+    "тип:текстом": {
+      "train": 1131,
+      "holdout": 80
+    },
+    "тип:числом": {
+      "train": 6653,
+      "holdout": 730
+    },
+    "тип:датой": {
+      "train": 1035,
+      "holdout": 169
+    },
+    "тип:деньгами": {
+      "train": 4568,
+      "holdout": 483
+    },
+    "тип:признаком": {
+      "train": 3532,
+      "holdout": 421
+    },
+    "тип:состоянием": {
+      "train": 1972,
+      "holdout": 744
+    },
+    "тип:номинальный": {
+      "train": 633,
+      "holdout": 44
+    },
+    "морфизм:если": {
+      "train": 1186,
+      "holdout": 322
+    },
+    "морфизм:то": {
+      "train": 1186,
+      "holdout": 322
+    },
+    "морфизм:из": {
+      "train": 786,
+      "holdout": 302
+    },
+    "морфизм:в": {
+      "train": 786,
+      "holdout": 302
+    },
+    "по закону": {
+      "train": 396,
+      "holdout": 539
+    },
+    "принимает": {
+      "train": 6451,
+      "holdout": 715
+    },
+    "возвращает:строку": {
+      "train": 393,
+      "holdout": 150
+    },
+    "возвращает:число": {
+      "train": 1584,
+      "holdout": 209
+    },
+    "возвращает:дату": {
+      "train": 396,
+      "holdout": 21
+    },
+    "возвращает:деньги": {
+      "train": 2896,
+      "holdout": 274
+    },
+    "возвращает:признак": {
+      "train": 1182,
+      "holdout": 61
+    },
+    "начинает с": {
+      "train": 6451,
+      "holdout": 715
+    },
+    "правило": {
+      "train": 6451,
+      "holdout": 715
+    },
+    "свойство": {
+      "train": 3690,
+      "holdout": 569
+    },
+    "пример": {
+      "train": 6451,
+      "holdout": 715
+    },
+    "правило:если": {
+      "train": 6451,
+      "holdout": 715
+    },
+    "правило:и": {
+      "train": 397,
+      "holdout": 20
+    },
+    "то добавить": {
+      "train": 4480,
+      "holdout": 483
+    },
+    "то результат": {
+      "train": 1971,
+      "holdout": 232
+    },
+    "свойство:результат": {
+      "train": 3690,
+      "holdout": 569
+    },
+    "дано(пример)": {
+      "train": 6451,
+      "holdout": 715
+    },
+    "ожидается результат": {
+      "train": 6451,
+      "holdout": 715
+    },
+    "сравн:не меньше": {
+      "train": 5665,
+      "holdout": 547
+    },
+    "сравн:не больше": {
+      "train": 2108,
+      "holdout": 233
+    },
+    "сравн:не равен": {
+      "train": 591,
+      "holdout": 159
+    },
+    "сравн:не равна": {
+      "train": 323,
+      "holdout": 45
+    },
+    "сравн:не равно": {
+      "train": 267,
+      "holdout": 25
+    },
+    "сравн:равен": {
+      "train": 7636,
+      "holdout": 1039
+    },
+    "сравн:равна": {
+      "train": 4672,
+      "holdout": 433
+    },
+    "сравн:равно": {
+      "train": 2703,
+      "holdout": 370
+    },
+    "сравн:больше": {
+      "train": 127,
+      "holdout": 134
+    },
+    "сравн:меньше": {
+      "train": 1190,
+      "holdout": 59
+    },
+    "операнд:результат": {
+      "train": 393,
+      "holdout": 21
+    },
+    "операнд:процентов от поля": {
+      "train": 1650,
+      "holdout": 204
+    },
+    "операнд:процента от поля": {
+      "train": 1316,
+      "holdout": 112
+    },
+    "операнд:процент от поля": {
+      "train": 395,
+      "holdout": 20
+    },
+    "операнд:поле": {
+      "train": 524,
+      "holdout": 30
+    },
+    "теорема:дано": {
+      "train": 1185,
+      "holdout": 453
+    },
+    "в данных": {
+      "train": 1185,
+      "holdout": 453
+    },
+    "по морфизму": {
+      "train": 596,
+      "holdout": 287
+    },
+    "затем по морфизму": {
+      "train": 697,
+      "holdout": 296
+    },
+    "применить морфизм": {
+      "train": 589,
+      "holdout": 166
+    },
+    "затем применить морфизм": {
+      "train": 304,
+      "holdout": 147
+    },
+    "следовательно": {
+      "train": 793,
+      "holdout": 175
+    },
+    "получаем": {
+      "train": 392,
+      "holdout": 153
+    },
+    "скаляр:да": {
+      "train": 4781,
+      "holdout": 637
+    },
+    "скаляр:нет": {
+      "train": 3569,
+      "holdout": 309
+    },
+    "скаляр:ничто": {
+      "train": 390,
+      "holdout": 23
+    },
+    "скаляр:число": {
+      "train": 6451,
+      "holdout": 715
+    },
+    "скаляр:отрицательное число": {
+      "train": 394,
+      "holdout": 20
+    },
+    "скаляр:дробное число": {
+      "train": 1305,
+      "holdout": 325
+    },
+    "имя:«ёлочки»": {
+      "train": 8996,
+      "holdout": 1499
+    },
+    "имя:голое": {
+      "train": 2369,
+      "holdout": 380
+    },
+    "комментарий //": {
+      "train": 5058,
+      "holdout": 828
+    }
+  },
+  "constructs_missing_in_train": [],
+  "constructs_missing_in_holdout": []
+}