Publish security_ir 384D development checkpoint 20261002-384-development-v1

README.md

@@ -0,0 +1,62 @@
+---
+tags:
+- autoencoder
+- formal-logic
+- experimental
+- ipfs-datasets-py
+---
+# security-ir: experimental 384D checkpoint
+
+**Held-out exact IR reconstruction: 0/2. Native structural validity: 2/2.** These two authored paraphrase examples share semantic targets with training; they are not independent real-world benchmark tasks. The model generates plausible but semantically incorrect candidates and must not authorize actions.
+
+This is an opt-in development artifact. It is **not the default symbolic compiler**,
+not a verified formalizer, and not a security repair policy. All qualification,
+admission and proof-authority flags remain false. No LLM/provider fallback runs.
+
+## Runtime
+
+Input embeddings use `thenlper/gte-small` at commit
+`17e1f347d17fe144873b1201da91788898c639cd` (384 dimensions). The embedding encoder
+weights are not included. Source inference requires the corresponding verified
+local encoder assets and rejects inputs above 512 tokens without truncation.
+The installed `ipfs_datasets_py` implementation must match the source hashes in
+the checkpoint; incompatible sources are rejected. This development runtime is
+currently implemented in the matching workspace checkout, not claimed to be in
+an existing PyPI release. No repository-supplied Python is executed on download.
+
+```python
+from ipfs_datasets_py.logic.security_ir import open_autoencoder
+model = open_autoencoder()  # library catalog pins a complete Hub commit + SHA256
+report = model.infer([{"id": "sample", "source_text": source, "embedding": vector384}])
+# Or: model.infer_texts([source], snapshot_path=verified_gte_snapshot)
+```
+
+## Training and scope
+
+Security, Intent and UI/UX heads inherit all compatible nonlexical tensors from
+a trained local Legal 384D head. New lexical rows use trained parent row means;
+no random parameters remain in their initial transferred state. Each native head
+used 12 authored training rows, 2 tuning rows, 500 epochs / 1000 optimizer updates;
+selection used tuning loss only. Outputs are explicitly local typed fragments,
+not complete verified programs or native documents. Security operator fragments
+still require their enclosing program and source-reference context. Legal retains
+its separate sparse core and learned formula head, with parser-derived features.
+
+These checkpoints were **not trained on the complete CVEFixes or SkillCenter
+corpora**. The separate full-feature structural-family experiment (15/9/8/6 routes)
+does not measure this 384D source decoder and is not its training loss or coverage.
+Weight/conditioning ablations demonstrate numerical dependence, not correct meaning.
+
+## Provenance and license
+
+The manifest records the actual parent hashes, training manifests, embedding
+assets and measured validation. Source screening found no private documents,
+credentials, benchmark solutions or third-party corpus text in the new authored
+fixture export. Original Legal files and the upstream Hugging Face datasets were
+not overwritten. This release makes no new license grant for inherited weights.
+The runtime source repository retains its AGPL-3.0 license; the new authored
+fixture export separately declares Apache-2.0. A corpus license is not asserted
+to be the weight license.
+
+The checkpoint and evidence are immutable under `releases/20261002-384-development-v1/`.
+Manifest SHA256: `d9fc241c7c6a039068681fab18b6a36fbcbd1752c2683ba1dc16423ce94aba2a`.

releases/20261002-384-development-v1/README.md

@@ -0,0 +1,62 @@
+---
+tags:
+- autoencoder
+- formal-logic
+- experimental
+- ipfs-datasets-py
+---
+# security-ir: experimental 384D checkpoint
+
+**Held-out exact IR reconstruction: 0/2. Native structural validity: 2/2.** These two authored paraphrase examples share semantic targets with training; they are not independent real-world benchmark tasks. The model generates plausible but semantically incorrect candidates and must not authorize actions.
+
+This is an opt-in development artifact. It is **not the default symbolic compiler**,
+not a verified formalizer, and not a security repair policy. All qualification,
+admission and proof-authority flags remain false. No LLM/provider fallback runs.
+
+## Runtime
+
+Input embeddings use `thenlper/gte-small` at commit
+`17e1f347d17fe144873b1201da91788898c639cd` (384 dimensions). The embedding encoder
+weights are not included. Source inference requires the corresponding verified
+local encoder assets and rejects inputs above 512 tokens without truncation.
+The installed `ipfs_datasets_py` implementation must match the source hashes in
+the checkpoint; incompatible sources are rejected. This development runtime is
+currently implemented in the matching workspace checkout, not claimed to be in
+an existing PyPI release. No repository-supplied Python is executed on download.
+
+```python
+from ipfs_datasets_py.logic.security_ir import open_autoencoder
+model = open_autoencoder()  # library catalog pins a complete Hub commit + SHA256
+report = model.infer([{"id": "sample", "source_text": source, "embedding": vector384}])
+# Or: model.infer_texts([source], snapshot_path=verified_gte_snapshot)
+```
+
+## Training and scope
+
+Security, Intent and UI/UX heads inherit all compatible nonlexical tensors from
+a trained local Legal 384D head. New lexical rows use trained parent row means;
+no random parameters remain in their initial transferred state. Each native head
+used 12 authored training rows, 2 tuning rows, 500 epochs / 1000 optimizer updates;
+selection used tuning loss only. Outputs are explicitly local typed fragments,
+not complete verified programs or native documents. Security operator fragments
+still require their enclosing program and source-reference context. Legal retains
+its separate sparse core and learned formula head, with parser-derived features.
+
+These checkpoints were **not trained on the complete CVEFixes or SkillCenter
+corpora**. The separate full-feature structural-family experiment (15/9/8/6 routes)
+does not measure this 384D source decoder and is not its training loss or coverage.
+Weight/conditioning ablations demonstrate numerical dependence, not correct meaning.
+
+## Provenance and license
+
+The manifest records the actual parent hashes, training manifests, embedding
+assets and measured validation. Source screening found no private documents,
+credentials, benchmark solutions or third-party corpus text in the new authored
+fixture export. Original Legal files and the upstream Hugging Face datasets were
+not overwritten. This release makes no new license grant for inherited weights.
+The runtime source repository retains its AGPL-3.0 license; the new authored
+fixture export separately declares Apache-2.0. A corpus license is not asserted
+to be the weight license.
+
+The checkpoint and evidence are immutable under `releases/20261002-384-development-v1/`.
+Manifest SHA256: `d9fc241c7c6a039068681fab18b6a36fbcbd1752c2683ba1dc16423ce94aba2a`.

releases/20261002-384-development-v1/manifest.json

@@ -0,0 +1 @@
+{"checkpoint_file":"checkpoint.json","dimension":384,"domain_id":"security_ir","embedding":{"dimension":384,"model_id":"thenlper/gte-small","revision":"17e1f347d17fe144873b1201da91788898c639cd"},"files":{"authored-corpus.json":{"bytes":180619,"sha256":"db7746b86c1149785a23802d545a9fdf329e209b175c625c8ce4e89d1361eb64"},"checkpoint.json":{"bytes":597741,"sha256":"855b3d6d85bb59e165930785bd49711df0fcc8fbc8e11329e2f4b65125520441"}},"proof_authority":false,"provenance":{"authored_fixture_license":"Apache-2.0","corpus_scope":"authored controls, known semantic targets, no CVEFixes/SkillCenter corpus training","corpus_sha256":"db7746b86c1149785a23802d545a9fdf329e209b175c625c8ce4e89d1361eb64","evaluation_limits":["paraphrase controls share semantic targets with training","no heldout IR grammar shape claim","not a full Publicus corpus training export","not formal correctness or task success evidence"],"existing_legal_artifacts_overwritten":false,"new_weight_license_grant":false,"parent_checkpoint_sha256":"969461ab82a2806e54ad33ba242a1eb62d032fa1b3cfc808b77c66dcc965aa62","private_or_credential_data_found":false,"release_scope":"public opt-in development artifact","runtime_source_license":"AGPL-3.0","source_embeddings":{"assets":[{"bytes":190,"name":"1_Pooling/config.json","sha256":"4be450dde3b0273bb9787637cfbd28fe04a7ba6ab9d36ac48e92b11e350ffc23"},{"bytes":583,"name":"config.json","sha256":"00273cdfb3e05c7304d180270bdef8adae45023e7bdfc396338bdd815943dc59"},{"bytes":66746168,"name":"model.safetensors","sha256":"9a1eb90bbac323ea08aa5629b624fe6ae75db121b904799c2266a1e2c2de22d2"},{"bytes":385,"name":"modules.json","sha256":"73348a3d8f8ae4cc5f851dc6b98be3ea819390b814e83656d5eebbadf75fe43e"},{"bytes":57,"name":"sentence_bert_config.json","sha256":"948201d8329907aae938fa62f9ceeed53f5694dacc2b87b9f3b78b37ee986529"},{"bytes":125,"name":"special_tokens_map.json","sha256":"b6d346be366a7d1d48332dbc9fdf3bf8960b5d879522b7799ddba59e76237ee3"},{"bytes":711661,"name":"tokenizer.json","sha256":"da0e79933b9ed51798a3ae27893d3c5fa4a201126cef75586296df9b4d2c62a0"},{"bytes":394,"name":"tokenizer_config.json","sha256":"e1790949631401af1bfb6c9c7aeec7fcf612e274d73579d99f704faea40c8ba7"},{"bytes":231508,"name":"vocab.txt","sha256":"07eced375cec144d27c900241f3e339478dec958f92fddbc551f295c992038a3"}],"device":"cpu","dimension":384,"dtype":"float32","elapsed_seconds_including_load":4.257302220910788,"model_id":"thenlper/gte-small","normalized":true,"revision":"17e1f347d17fe144873b1201da91788898c639cd","truncated":false,"vectors_executed":48},"third_party_corpus_text_included":false,"training_manifest":[{"embedding_sha256":"026cbbabf3973e9e11f5ddceaecc704be3572d162bf00da8c6be441e2c64ec2c","id":"security_ir:train:0","source_sha256":"833556db828032d49efc4c58a380411c53465c917da89bf613a607a10cee2ba6","target_sha256":"2db133a15af147d41db6386953e50ef760e8b009c0ba2b4ef91c140874f05480"},{"embedding_sha256":"6d5a7d8471ba8d203978e4b77435a8c24bc493f76c680be35a8c1bd31bdd7403","id":"security_ir:train:1","source_sha256":"ccc2b65889226c357d6c5bda5544dfc34b7488ba12a798bcbfd695cddcb98cf8","target_sha256":"350251014cd6db588df0b6da57ecf58beea52cf36fe8c4366ea438c28afea4b4"},{"embedding_sha256":"dc9b9a5b3c32a7cedfc8a267e976267af798839ed3c28fae3a8e4e89f3fc9223","id":"security_ir:train:2","source_sha256":"5fe8ec022167ca0e5b8912fc343071d762c4c1160b045ec441c07de704048303","target_sha256":"0d8a3c27e83a2dad735225fa72e262460f2187c428ce4f3c6a308265b09b3492"},{"embedding_sha256":"aa1bfa976bd7e5a51a44cd89fa5fb95459e190f17b331b04120e8e9d73d87493","id":"security_ir:train:3","source_sha256":"2967eacb633a8b98c667529b901722afd86cb51cb103fe779b97489701289227","target_sha256":"476b25a8db7849cce7099bd08c5509ab0b3ac088fb4e5ca77bf9c1179cc9d218"},{"embedding_sha256":"dc1264b5f5e65672455e95994064152833b1338717cbd84dc75d2612d6cb60bd","id":"security_ir:train:4","source_sha256":"d5aceec1160ac82aa38fa5e52030e520588d662c4bdf72a72f5a1f336a51a8af","target_sha256":"1e85a7e84aea0bf2b5bc3d0ca9ac35c508724272024147bd4017ca1a7820fbfc"},{"embedding_sha256":"0179762fb06b32afa0d32f50e8012ba431a55d3b7c35a75a4744397c9976a01f","id":"security_ir:train:5","source_sha256":"f056a5c735a77c53a2ee3a216aba5a81df57364dfff88bd30872ce7d8fd1a5bd","target_sha256":"e7c02097dce5173f14aa6ce449503a835bd02b30e323de2e112534cb78782c8c"},{"embedding_sha256":"144cb03552217461cdfd7fd0b9e53e047be78d908193c57707c33fb537e7e681","id":"security_ir:train:6","source_sha256":"bc666905c1f2bcd4a701abbe559f1f91784132decd6ba46f18887171bd4550a9","target_sha256":"edcfd2a0bff80075cc7541b9961aaf3a4c36d7077121eeb7ef9100e40794c502"},{"embedding_sha256":"d18533ad2ff86852f5a9cbf2d0e33edca42c61ef124ae0bacaa3123f16bc5086","id":"security_ir:train:7","source_sha256":"ec2d05f06ebf8506f14ebec38c3ea8d885ea932980692ee85524c90f1bfa74cc","target_sha256":"53b85297b618ed83361f6577ea4125e2db23863ccc3931a025e27b78306ecac8"},{"embedding_sha256":"c63d6ea6f9f23ed379a3f78c268faa65aa1850feb1ebf65be65dd99680961d13","id":"security_ir:train:8","source_sha256":"d172a489cc3abbe644164bdb783fc793f20c3d042cab5023dd4a641764685c0b","target_sha256":"4978838fd2ef0a9d59c0b887f0b5544363a9b5385ffeadecb81c4e210183927b"},{"embedding_sha256":"6fcec1dc314cf6df1809cfd83aaa84aef3b7b4170088cff58f71cee6bf899159","id":"security_ir:train:9","source_sha256":"a8db6d79041ec15e9fca23b7197df9396825dc2b33ba48b13546fc634c329403","target_sha256":"3d7f961981516b87f249142e3cec4f36228386174403af46ab4f4b3b235ba7e5"},{"embedding_sha256":"dcbe92dc82ebd773e14ed7983bbd1bfb30b324c5e93827a03195233ba958e410","id":"security_ir:train:10","source_sha256":"6c48fed71cbb99753d7cd3206c26fb0174a1dc512203415093cccca11276310e","target_sha256":"f76c05269d810a378a5eceda615a6a43eb0939335b7a36a3a5b041aac304582e"},{"embedding_sha256":"19ba32ed6899be97e7098e7cecddc4212a568174789d1c117053ca4a35dfa026","id":"security_ir:train:11","source_sha256":"c1917f28817fcb4cea28d1f3afe2192080332729b40bd87aa0d6b3996995faf5","target_sha256":"2a2f0accb542e291c0fdd7a4f0f9fa784352bd4cc84d074b2875e1c6e6ca5a43"}],"transfer":{"exact_inherited_tensors":["condition.bias","condition.weight","decoder.bias_hh_l0","decoder.bias_ih_l0","decoder.weight_hh_l0","decoder.weight_ih_l0","projection_down.bias","projection_down.weight","projection_up.bias","projection_up.weight"],"initial_state_sha256":"919191a48d31d6e3b36a4a8f9ec83f0be025131422a70c2addc7a78f1dd87bce","lexical_mapped_tensors":["output.bias","output.weight","target_embedding.weight"],"new_token_initialization":"trained_parent_lexical_row_mean","parent_modified":false,"random_parameters_used":false},"validation_manifest":[{"embedding_sha256":"b06d2d8c16f55ec1f924a0829e516b9477eae6fbebb20eb4e7d92b704a667ebf","id":"security_ir:validation:12","source_sha256":"e65dd07a9db1cdf9ef989a996377b34bcc612523f8f324ed8f5c189200b63b01","target_sha256":"2db133a15af147d41db6386953e50ef760e8b009c0ba2b4ef91c140874f05480"},{"embedding_sha256":"f48784614f06120a527f9651c81caf09b6bacfba0652268d1a81df114217eb92","id":"security_ir:validation:13","source_sha256":"cd34129bba788c02b1db1b7c3327779f0db78a4ce1033f45157d7527d3fb2efb","target_sha256":"476b25a8db7849cce7099bd08c5509ab0b3ac088fb4e5ca77bf9c1179cc9d218"}]},"release_stage":"development","runtime":"domain_384_v1","schema":"ir-384-hub-package/v1","validation":{"default_symbolic_formalizer_replaced":false,"exact_heldout_targets":0,"formal_correctness_proven":false,"heldout_count":2,"inference_did_not_mutate_weights":true,"metrics":{"normal":{"count":2,"exact_targets":0,"generation_changed":0,"reconstruction_changed":0,"valid_candidates":2},"shuffled_embedding":{"count":2,"exact_targets":0,"generation_changed":2,"reconstruction_changed":2,"valid_candidates":2},"zero_condition":{"count":2,"exact_targets":0,"generation_changed":2,"reconstruction_changed":0,"valid_candidates":0},"zero_decoder":{"count":2,"exact_targets":0,"generation_changed":2,"reconstruction_changed":0,"valid_candidates":0},"zero_projection":{"count":2,"exact_targets":0,"generation_changed":2,"reconstruction_changed":2,"valid_candidates":0}},"native_fragments_only":true,"proof_authority":false,"reload_parity":true,"source_fidelity_proven":false,"source_quality_accepted":false,"valid_heldout_targets":2}}