Complete catalog recovery and enrichment

v11.5 powered re-run: effective n=28, full=1/28 [0.006,0.177], physchem=0/28 [0.000,0.121], verdict=underpowered

v11.5 enrichment: VP/BP 67→340, structure 1731→1741, eval-pairs newly representable=3, leak-rejected=0, formula/MW-rejected=15

v11.3 eval harness: remove unused profile lookup import

v11.3 eval set expanded: n=83 authoritative pairs, shape distribution=41/39/3

fix(v11): preregister mode-aware retrieval evaluation

v11 material profile table: available fields, N=557, deferred enrichments

pimt v10 add line-drawn training assessment runner

pimt v9 allow hf registry db path

pimt v9 require registry db for gbm reproduction

pimt v9 substantivity gbm train publish path

pimt v8 diagnostics and qa artifacts

pimt v8 substantivity gate and pubchem cache

Precommit PIMT v8 multiseed split seeds

pimt v8 run statistical validation

pimt v8 statistical validation tests

character verdict: discriminative-weak under distributional metric

character distributional metric: cosine/JS over prevalence vectors, mean=0.080

character diagnostic: artifact-fixes — prevalence-by-genre, threshold-sensitivity, retract fake best-grouping

character 1.000 diagnosis: artifact — superclass sets identical

character grouping diagnostic: 26-class J=0.793, 5-superclass J=1.000, recommend pause labels

phase1 recompute: drop-10 ρ=0.708, prefer-measured ρ=0.663, baseline survived Y, noise floor=0.479

Add Fraterworks ingestion and upload data split support

feat: add WiseMoor v9 provenance dataset

feat: harden v7 HF training and eval

feat: prepare v7 HF training and evaluation

chore: add model audit run artifacts

feat: add Poucher substantivity targets

feat: expand descriptor target coverage

chore: remove temporary diagnostic scripts

feat: use AI (OpenRouter) to classify GoodScents prose into perfume tiers

feat: fetch GoodScents substantivity data and calibrate thresholds

feat: extract perfumer tier labels from Poucher Vol II odour classification tables

Update HF job script for v6: dataset v8, checkpoint pimt_v6.pt, scalar_stats

v6: ConcentrationAwarePyramidHead + OAV fix + curated descriptors + integration

fix: monkey-patch is_xet_available to force legacy LFS upload in HF job

fix: more aggressive Xet bypass for v5 checkpoint upload

feat: 3-tier scent pyramid objective head + v4 dataset + v5 training job

persist shalimar extracted text and update predict script path

hf_train_job: monkey-patch XetProgressReporter to avoid upload crash

hf_train_job: uninstall hf_xet before upload to avoid Xet progress-reporter crash

hf_train_job: disable Xet to avoid progress-reporter bug on large checkpoint upload

hf_train_job: install huggingface_hub before downloading source/dataset

hf_train_job: use snapshot_download (v1/v3 working pattern)

hf_train_job: lazy-import huggingface_hub to match v3 pattern

hf_train_job: add upload diagnostics and skip guard

hf_train_job: revert to mount-based writable registry pattern from v3

hf_train_job: snapshot_download source+data, writable registry

Lazy-load embedding engine; force HF_REGISTRY_PATH to /data/registry.db

hf_train_job: persistent diagnostic log + registry path tracing