Add files using upload-large-folder tool

.gitattributes

@@ -1,35 +1,4 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
 *.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz 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
+*.gguf        filter=lfs diff=lfs merge=lfs -text
+*.jsonl       filter=lfs diff=lfs merge=lfs -text
+*.pyc         filter=lfs diff=lfs merge=lfs -text

DATA-CARD-training-corpus.md

@@ -0,0 +1,141 @@
+# Data card — PyBytecode v2 / v3 training corpus
+
+The corpus behind `pybytecode-v2-1.5b` and `pybytecode-v3-1.5b`. Written 2026-08-04; no data card
+existed before.
+
+**This corpus is not distributed.** Section 5 states why, as a property of the artifact rather
+than an apology. The model weights are unaffected — see `WEIGHTS-LICENSE-PROPOSAL.md`.
+
+---
+
+## 1. What it is
+
+| | |
+|---|---|
+| Task | Python 3.12 bytecode disassembly → original source |
+| Rows | **48,196** (identical row set in v2 and v3) |
+| Row shape | `{"input": <disassembly>, "output": <source>}` |
+| Source dataset | `codeparrot/github-code-clean` (the dataset itself is Apache-2.0) |
+| Shards used for training | 0–5 |
+| Unit | one top-level function plus its transitive helpers and imports |
+| Python | 3.12, `optimize=0` |
+| Files | `data/foundry/pybytecode-v2_train.jsonl`, `-v3_train.jsonl`, `-v3-sft_train.jsonl` |
+
+v3 differs from v2 in the **input representation only**. v2's `rep.py` omitted the exception
+table's `end`, so a bare `try:` body and a `try/else:` body that compile to the same instruction
+stream were byte-identical in the model's input. v3 emits `EXC try=Ls..Le -> ...` and the end
+label joins the label set, making the two distinguishable. 14,122 of 48,196 inputs changed;
+no row was added, removed or relabelled (`data/foundry/pybytecode-artifacts/v3_build_report.json`).
+Any v2→v3 delta is therefore attributable to the representation fix alone.
+
+## 2. Licence filtering — what was dropped, and why
+
+`scripts/pybytecode/extract_v2.py` filters **per row** on `github-code-clean`'s `license` column
+and keeps seven permissive values:
+
+```
+mit   apache-2.0   bsd-2-clause   bsd-3-clause   isc   unlicense   cc0-1.0
+```
+
+Every GPL, LGPL, AGPL, MPL and EPL row is dropped before extraction, along with everything the
+column does not positively identify. The intent was to avoid training on reciprocally-licensed
+source; the filter is a hard gate, not a preference.
+
+Rows are then dropped by a chain of quality gates, each counted rather than silently applied:
+unparseable; not stable under `ast.unparse` round-trip (`canonicalise(canon) != canon`); does not
+compile; disassembly or source over the size ceiling; duplicate of an already-kept unit
+(SHA-1 of the canonical source).
+
+## 3. Decontamination against the test sets
+
+Three layers, because one is never enough (`scripts/pybytecode/build_final.py`):
+
+1. **Shard-disjoint** — training from shards 0–5, held-out pool from shards 8–9. The held-out
+   pool was never read during training extraction.
+2. **Repo-disjoint** — any repo appearing in training is removed from the held-out pool outright.
+   A repo can span shards, so layer 1 does not imply this.
+3. **Fingerprint-disjoint** — an identifier-blind structural fingerprint (every `Name`/`arg`/
+   attribute → placeholder, every literal → its type name). This catches the same algorithm under
+   renamed variables, which exact match misses. It caught 15 items exact match missed.
+
+The CSN benchmarks apply the same identifier-blind fingerprint against this corpus. On the
+600-row licensed rebuild it removed 3 rows.
+
+## 4. Labelling
+
+Labels are the canonical source itself, so the supervision is exact by construction. The grading
+tier attached to each unit was assigned **by running an oracle**, not by inspection:
+
+- **tier A / behavioural** — differential execution. Kills 100% of injected semantic bugs; the
+  gold tier.
+- **tier B / stub** — stubbed execution. Kills 78.9%, so scores on it are an **upper bound**.
+- **tier C / AST-exact** — undercounts by roughly 2.4×, so scores on it are a **lower bound**.
+
+Only 5.43% of all real top-level Python functions survive to be behaviourally adjudicable. That
+ceiling is why the byte-identical recompile oracle exists: it needs no runnable environment and
+so has 100% coverage.
+
+## 5. The corpus cannot be redistributed
+
+**Per-row attribution was not retained.** `extract_v2.py` carried `repo` and `license` on every
+row through extraction, and `build_final.py:118` writes only `{"input", "expected"}` when it
+emits the final splits. The intermediate pool that still held the metadata was written to `/tmp`
+and no longer exists.
+
+The consequence is specific: the corpus is 48,196 excerpts of MIT-, BSD-, Apache-, ISC- and
+public-domain-licensed source, and every one of those licences except the two public-domain
+dedications requires the copyright notice to be reproduced with the copy. We cannot produce those
+notices, because we no longer know which row came from which repository. Redistributing the file
+would strip required notices from tens of thousands of copyright holders.
+
+This is a limitation of the artifact, not of the licences: nothing about the corpus is unlicensed
+or reciprocally licensed. It is not fixable by adding a licence file, and it is not repaired by
+listing the source dataset — attribution under these licences is per-work, not per-collection.
+
+**It is fixable by rebuilding.** Extraction is deterministic and the source dataset is public.
+Carrying `repo`, `license` and file path through `build_final.py`'s writer — one line — produces
+an equivalent corpus that *is* redistributable with a `NOTICES` file. That is the recommended
+fix for a v4 and it does not require retraining anything to be useful.
+
+## 6. Held-out evaluation sets built from the same corpus
+
+`data/foundry/pybytecode-v3-ood-{behavioural,stub,ast,doc}_test.jsonl` are drawn from the
+held-out shards of this same corpus and **inherit section 5 exactly** — they carry no attribution
+and cannot be redistributed either.
+
+They are named `-ood-` but they are **not out-of-distribution**: they are the same source and the
+same distribution as training, held out three ways. Matching all 974 canonicalised
+`google-research-datasets/mbpp` rows against them yields **0 matches** in all four files, while
+the v1-era `pybytecode-ood_test.jsonl` matches at **400/400**. Only that v1-era file is MBPP.
+Full evidence in `LICENSING-DETERMINATION.md` §4.
+
+They remain sound *generalisation* tests, and the numbers measured on them stand. Only the label
+is wrong, and it should be corrected wherever it appears.
+
+**On the tier-A set specifically (n=279, the set behind the 91.04% and 97.49% figures):** it is
+the complete behaviourally-adjudicable population of the held-out pool — the 400-row cap did not
+truncate it, whereas the stub and ast tiers both hit that cap. The pool was therefore larger than
+400 units, but its exact size is **not recoverable**: `build_final.py` printed its census to
+stdout and no run log was kept. The framing "279 of 400 = 70% of the set" does not describe these
+files (the four tiers are largely disjoint populations — the 279 behavioural rows share 9 function
+names with the 400 stub rows and 8 with the 400 ast rows) and should not be used. Report the
+denominator as 279 with the sentence above, or rebuild the pool to recover the true rate.
+
+## 7. Known confound, stated rather than engineered away
+
+`extract_v2.py` carries `from __future__ import annotations`, and its `compile()` call inherits
+the flag, so the training bytecode has PEP-563 stringised annotations. **Real `.pyc` files are not
+compiled that way.** The benchmarks deliberately compile with `dont_inherit=True`, giving the
+model an input distribution it was not trained on for annotated functions. Any resulting handicap
+is a real property of the model and is counted against it.
+
+## 8. Provenance summary
+
+| Field | Value |
+|---|---|
+| Base model | `Qwen/Qwen2.5-Coder-1.5B-Instruct` (Apache-2.0) |
+| Rows | 48,196 |
+| Method | LoRA r=16, α=32, all attention + MLP projections, 1 epoch, lr 2e-4 |
+| Build scripts | `scripts/pybytecode/{extract_v2,build_final,build_v3,rep,gen,verify}.py` |
+| Grader hashes | all six match `grader.sha256` in `data/models/models.jsonl` |
+| Redistributable | **No** — §5 |

LICENSE

@@ -0,0 +1,202 @@
+
+                                 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.

NOTICE

@@ -0,0 +1,24 @@
+PyBytecode
+Copyright 2026 Blazing Customs
+
+This product includes software developed from Qwen2.5-Coder-1.5B-Instruct
+(https://huggingface.co/Qwen/Qwen2.5-Coder-1.5B-Instruct), Copyright Alibaba Cloud,
+licensed under the Apache License, Version 2.0.
+
+The upstream repository ships a LICENSE file and no NOTICE file (checked 2026-08-04 against
+the Hugging Face API), so there is no upstream NOTICE text to append here.
+
+CHANGES MADE TO THE LICENSED WORK, as required by Apache-2.0 section 4(b):
+
+  The base model was fine-tuned with LoRA (rank 16, alpha 32, applied to q_proj, k_proj,
+  v_proj, o_proj, gate_proj, up_proj and down_proj; 1 epoch; learning rate 2e-4) on 48,196
+  pairs of Python 3.12 bytecode disassembly and the source that produced it. The resulting
+  adapter was merged into the base weights. No architecture, vocabulary or tokenizer change
+  was made.
+
+The training corpus is not distributed. Its per-row attribution was not retained, so it cannot
+be redistributed without stripping required notices from its upstream authors. See
+DATA-CARD-training-corpus.md. This constrains the corpus only; it does not encumber these
+weights, which are trained parameters rather than a copy of any source text, and which were
+derived from a corpus pre-filtered to seven permissive licences with all GPL, LGPL, AGPL, MPL
+and EPL rows dropped before extraction.

ORACLE-LIMITS.md

@@ -0,0 +1,133 @@
+# The oracle's real limits
+
+The verifier is the reason to use PyBytecode at all, so its limits belong in front of a user, not
+in an appendix. Everything here is measured; sources are named per section.
+
+---
+
+## 1. The pre-flight 100% proves almost nothing. Read this before quoting it.
+
+Every grading command prints `PRE-FLIGHT 600/600 = 100%` before it scores. **That number is
+trivial by construction and is not evidence of soundness.**
+
+Pre-flight grades each reference label against itself. The oracle asks whether
+`compile(prediction)` and `compile(reference)` produce the same code object — so at pre-flight it
+is comparing `compile(x)` with `compile(x)`. It would return 100% for *any* deterministic
+function of the source, including a stub that hashes the input string and ignores the bytecode
+entirely.
+
+What pre-flight actually detects is a **broken harness**: a benchmark whose `.pyc` files do not
+match their sources, a Python version mismatch (3.11 or 3.13 against a 3.12 benchmark), a corrupt
+row. Those are real failure modes and worth catching, which is why it runs. But a passing
+pre-flight says the instrument is plugged in, not that it measures anything.
+
+**Soundness evidence comes from the mutation test and the blind-spot probes, not from
+pre-flight**: corrupt a label and require the oracle to reject it. Measured
+(`evidence/ORACLE-MUTATION.md`): **0 true survivors in 1,239 mutants**, and
+**18/18 targeted blind-spot probes** behave as required — including the historical failure where
+a `try:` body and a `try/else:` body were indistinguishable, docstring changes, docstring
+removal, float-vs-int, bool-vs-int and `-0.0` vs `0.0`.
+
+Even the mutation kill rate is weak evidence on its own: for a byte-identical oracle a kill is
+close to tautological, since a mutant survives only if it compiles to a structurally identical
+code object. The probes are the load-bearing test, because they ask the question that actually bit
+us once — *is a behaviourally load-bearing field missing from the fingerprint?*
+
+**Caveat on mutation supply, stated rather than hidden:** 188 of 600 wild rows (31%) produced no
+effective mutant within 30 tries, and 323 void attempts were discarded. The wild kill rate is
+measured on the 412 rows that did produce one.
+
+## 2. The 0.33% wild false-reject floor
+
+Against `.pyc` files built by someone else, the oracle refuses a small fraction of correct
+answers. Measured on 600 wild install-time `.pyc` from installed site-packages
+(`evidence/GATE-RESULT.md`):
+
+| | certified | false reject | false accepts |
+|---|---|---|---|
+| L0 (old constant encoding) | 585/600 = 97.5% | 15 = 2.5% | 0 / 1,274 |
+| **L1 (shipping)** | **598/600 = 99.67%** | **2 = 0.33%** | 0 / 1,274 |
+
+13 of the 15 L0 failures were our own defect — `repr()` of a `set`/`frozenset`/`dict` follows the
+compiling process's hash seed, which also made the L0 verdict **non-deterministic** (585 / 592 /
+584 / 585 / 589 under `PYTHONHASHSEED` 0–4). L1 fixes it and returns 598 under all five seeds.
+
+The remaining **0.33% is a real floor and is not fixable.** One distinct module
+(`pandas/_testing/__init__.py`) compiles differently under CPython 3.12.3 than under 3.12.13 —
+`co_code` 2,692 vs 2,696 bytes, and a differing `co_exceptiontable`. The source is correct; the
+*compiler patch release* differs. No normalisation removes this without abandoning the
+byte-identical guarantee.
+
+**It degrades to a false REJECT, never a false accept.** You are told "unknown" about a correct
+answer; you are never told "verified" about a wrong one. That is the safe direction, and it is the
+direction the design chose deliberately.
+
+## 3. Optimization level must match the producer's, or verification collapses
+
+A `.pyc` built with `-O` or `-OO` is a different code object. Measured on 679 sources compiled by
+a foreign interpreter at each level and graded at each level:
+
+| producer ↓ / grader → | 0 | 1 | 2 |
+|---|---|---|---|
+| **0** | **100.0%** | 95.43% | 23.86% |
+| **1** | 95.43% | **100.0%** | 24.15% |
+| **2** | 23.86% | 24.15% | **100.0%** |
+
+The diagonal is 679/679 at every level. Guessing wrong is **not** a graceful degradation — it
+collapses to ~24%. There are only three levels and trying all three costs three compiles, so this
+is *"needs normalisation to survive"*, not *"breaks"*. **The harness does not currently search the
+three levels automatically; a user verifying a foreign `.pyc` must do it.**
+
+**What trying all three costs you, and it is not nothing.** The certificate changes meaning from
+*"byte-identical to the code object the original source compiles to"* to *"byte-identical to the
+code object that was actually shipped"*. At `optimize>=1` docstrings are **absent from the `.pyc`**,
+and at `>=2` asserts are gone too. So against an `-O` artifact, **docstring recovery cannot be
+certified at all** — the information is not in the file. That bears directly on our headline
+differentiator: 115 of 679 benchmark rows carry a real docstring, and none of that could be proven
+against an `-O` `.pyc`. It is a limit of the artifact, not unsoundness in the oracle.
+
+## 4. What the oracle deliberately ignores
+
+Excluded from the fingerprint, with how often each would have caused a false reject on the 600
+wild rows had it been included:
+
+| Field | In fingerprint? | Rows differing / 600 |
+|---|---|---|
+| `co_filename` | excluded | **600 (100%)** — every wild `.pyc` carries its builder's absolute path |
+| `co_linetable` | excluded | **122 (20.3%)** — varies across builds with no semantic content |
+| `co_firstlineno` | excluded | 0 |
+
+Docstrings, `co_consts`, `co_names` and `co_exceptiontable` are **included** — the first because
+docstring fidelity is a claim we make, the last because omitting it once produced a false proof.
+
+## 5. Unverified means unknown, not wrong
+
+The oracle is **sound but incomplete**:
+
+```
+verified   = PROVABLY correct. Identical code object => identical behaviour. No false positives.
+unverified = UNKNOWN. A correct decompilation that compiles differently — a `while` where the
+             original had a `for`, a differently-ordered but equivalent boolean — does not verify.
+```
+
+Reported accuracy is therefore a **lower bound on correctness**, not an estimate of it. Treating
+the unverified remainder as errors understates the model; treating it as correct is unsafe.
+
+## 6. Not tested — unknown, not claimed
+
+- **Cross-minor (3.13).** No 3.13 interpreter on the measurement box; nothing was downloaded.
+  The benchmark and the model are 3.12 only.
+- **PyInstaller / Nuitka containers.** `import PyInstaller` → `ModuleNotFoundError`. Not measured.
+- **`.pyc` from non-CPython or patched builds.** Not measured. Given that a *patch release*
+  already produces the 0.33% floor, a patched build is a live risk, not a theoretical one.
+- **Obfuscated or deliberately adversarial bytecode.** Not measured. No malware was fetched.
+
+## 7. Where these limits are stated to users
+
+| Limit | Stated in |
+|---|---|
+| Pre-flight is trivial | this file; `harness/README.md`; both benchmark data cards |
+| 0.33% wild false-reject floor | this file; `weights/MODEL-CARD.md` |
+| `-O` mismatch collapse, docstrings unprovable | this file; `weights/MODEL-CARD.md` |
+| unverified ≠ wrong | this file; `weights/MODEL-CARD.md`; `harness/README.md` |
+| 3.13 / PyInstaller untested | this file; `weights/MODEL-CARD.md` |