Archive 4 checkpoints (1-4)

paper-releases/pbisc-diffusion/updates/20261012T130738Z-strictpairs-implementation/diffusion/dataset/server_1/model8/plans/paper_figures_strictpairs_20261012/archive/20261012T130738Z-strictpairs-implementation/snapshot/PLAN.md

@@ -0,0 +1,12 @@
+# Authorized PBISC-Diffusion Figure 1–5 implementation
+
+User authorized 2026-10-12. The full user-supplied plan in the chat is binding.
+
+1. Build exact sample-level support manifests shared by LoRA, paired calibration, scSemiProfiler and deconvolution. Healthy outer folds: nested n=5,10,20,40; seed20261012+fold. Disease: preserve five existing donor draws, choose earliest identity-passed matching-stratum visit per donor. Exclude all other selected-donor visits from train and test. Strict n has no extra validation sc; fixed3000 EMA; scSemi HVG uses n only.
+2. Figure1: full pretraining→exact5pair LoRA→real-bulk inference; corpus; masks; K_cond/depth/N_gen independently labelled; external ref-cal alternative.
+3. Figure2: held-out native pseudobulk architecture/panel ablations re-score P/C; healthy full/r4/r16/r64 existing5fold; new strict-n curve and external AIFI reference calibration.
+4. Figure3: strict5pair ImMicro OOF real-reference UMAP, canonical marker heatmap/dotplot, composition, P/C/C2ST/MMD/EN; operational compute excludes supplied pretraining.
+5. Figure4: exact5pair and original5fold disease comparisons; existing metric family; donor dependence; deconvolution same support.
+6. Figure5: paired rare states plus descriptive Behcet; zero-aware aggregation, actual donor IDs, common depth/cell budget, required markers, all states/draws, no false independent Behcet claim.
+7. Figure6: waiting for user-prepared bulk-only reproduction data, no work.
+8. Generate source data/provenance, vectorPDF/SVG/PNG and PPTX; integrate only supported Results/figures; versioned paper archive; no GitHub push or Overleaf synchronization.

paper-releases/pbisc-diffusion/updates/20261012T130738Z-strictpairs-implementation/diffusion/dataset/server_1/model8/plans/paper_figures_strictpairs_20261012/archive/20261012T130738Z-strictpairs-implementation/snapshot/receipts/prepare-strict-conditions.json

@@ -0,0 +1,7 @@
+{
+  "state": "completed",
+  "signature": "5d47d23bc2cbac7b068bc06022d7dc84d6525f257bb505207f57ac6b6f0401ee",
+  "outputs": [],
+  "seconds": 4.280744560994208,
+  "at": "2026-10-12T10:45:19.800668+00:00"
+}

paper-releases/pbisc-diffusion/updates/20261012T130738Z-strictpairs-implementation/diffusion/dataset/server_1/model8/plans/paper_figures_strictpairs_20261012/archive/20261012T130738Z-strictpairs-implementation/snapshot/receipts/verify-cpu-final-with-r.json

@@ -0,0 +1,14 @@
+{
+  "state": "completed",
+  "signature": "930c23c681af68a7ced60588fb66e21cdf8ffe78add280b18672417fa0e46741",
+  "outputs": [
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paper_figures_strictpairs_20261012/logs/verify-cpu-final-with-r.log",
+      "bytes": 690,
+      "mtime_ns": 1791810354733178176,
+      "sha256": "567e1c2e5e1fb762d93e9c88717eef96294b51afe330386c76b4b644734ac086"
+    }
+  ],
+  "seconds": 36.068204927956685,
+  "at": "2026-10-12T13:05:55.368065+00:00"
+}

paper-releases/pbisc-diffusion/updates/20261012T130738Z-strictpairs-implementation/diffusion/dataset/server_1/model8/plans/paper_figures_strictpairs_20261012/archive/20261012T130738Z-strictpairs-implementation/snapshot/receipts/verify-cpu-final.json

@@ -0,0 +1,14 @@
+{
+  "state": "completed",
+  "signature": "4eaea785e18694ef7a0710acbd3d41bae9a091c2e419c68eb51831e82654b6e6",
+  "outputs": [
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paper_figures_strictpairs_20261012/logs/verify-cpu-final.log",
+      "bytes": 701,
+      "mtime_ns": 1791810241507214043,
+      "sha256": "9a7bb6df0fba1bca7c449c3db5b3f379eb03df0c6f63d43f7d03e47f7f07d751"
+    }
+  ],
+  "seconds": 37.88490461395122,
+  "at": "2026-10-12T13:04:03.055000+00:00"
+}