Archive 4 checkpoints (1-4)

paper-releases/pbisc-diffusion/updates/20261015T100456Z-figure1-diffusion-readable-corpus/diffusion/dataset/server_1/model8/AGENTS.md

@@ -0,0 +1,100 @@
+# Model 8 PBISC-Diffusion paper maintenance
+
+This active Model 8 source and its experiments are part of the
+`pbisc-diffusion` paper. At task start, read
+`../../paper-release-20260928/maintenance/POLICY.md` along with this
+directory's README and applicable parent instructions.
+
+At the end of each related task, follow that policy's versioned update
+workflow for the exact source, experiment, checkpoint, or evidence assets that
+changed. From the release root, use
+`maintenance/update.py prepare --request FILE --id ID`, inspect the prepared
+update, then run `maintenance/update.py publish ID --execute` and
+`maintenance/update.py status` under the policy. Record a no-op or pending
+status when no
+asset changed or publication failed; never describe that as an upload. Do not
+regenerate or alter the frozen 2026-09-28 release, its root manifests, or the
+HF `paper-20260928` revision. Inference assets are location plus receipt by
+default unless the task explicitly needs their payload.
+
+The private dataset must stay private. Include datalake, VAE, or Model 7 assets
+only with exact evidence that they are dependencies of this diffusion paper.
+The parent host GPU safety rules still apply.
+
+## Reuse existing experiments before starting compute
+
+User instruction, 2026-10-13: avoid repeating completed Model 8 experiments.
+Start with `EXPERIMENT_INDEX.md` and query `EXPERIMENT_REUSE_INDEX.jsonl` using
+`scripts/audit_experiment_reuse.py`. A missing output in a new plan directory
+does not establish that an experiment has never been done.
+
+- Before training, generation, or deconvolution, compare existing source
+  receipts: base checkpoint, exact support/test sample and donor IDs, data and
+  condition content, measured genes/masks, validation access, optimization and
+  selection schedule, seed, and implementation/configuration provenance.
+- Registry matches are candidates, not automatic equivalence. Verify payloads
+  and write truthful adoption evidence. Preserve original execution provenance;
+  do not relabel an old artifact as newly executed strict work.
+- Changing a sampling seed or receipt format does not justify retraining an
+  existing model. Reuse verified checkpoints and run only the genuinely changed
+  operation. Missing native receipts call for adoption, not automatic reruns.
+- Record the candidate sources and the exact scientific reason for any new
+  operation before launching it. Unknown provenance requires investigation,
+  not a conclusion that new compute is necessary. Consult newer live receipts
+  when the inventory predates an ongoing campaign.
+- Distinguish existing results, CPU-only statistics/plotting, a new protocol
+  update, and missing data in plans and decks. Never label all four as an empty
+  TODO. Archive/backup copies and tensor-identical models are not independent
+  experimental evidence.
+
+The 2026-10-13 audit and confirmed duplicate-training proof are in
+`plans/paper_figures_strictpairs_20261012/audit_reuse_20261013/`.
+Sepsis deconvolution uses verified historical draw-specific outputs through
+`tools/adopt_deconv_legacy.py` in that plan; its methods must not be rerun as a
+fallback when adoption validation fails. Current expanded evidence master (156 slides, all 155 parent page IDs represented; Figure 1a clarified):
+`presentations/2026-09-30--paper-summary--r00/PBISC-Diffusion_expanded_master_20261015_figure1_focused.pptx`.
+Use its `data/figure1_focus_20261015/page_map.csv` and source manifest.
+Earlier 155/154/142/140/131-slide decks remain preserved source versions, not the current master.
+The 34-slide review is a historical subset, not the master. The approved method palette is
+`presentations/2026-09-30--paper-summary--r00/PBISC-Diffusion_color_palette_20261013.md`:
+PBISC LoRA #168FE5, calibrated #59A8DF, raw #C3E2F4; scSemiProfiler #83C6A8;
+BayesPrism #D9B23F and CIBERSORTx #F0D77D. Preserve biological/cohort colours and quantitative heatmap scales.
+Older gold/blue source figures are preserved historical assets; use the final semantic palette derivatives in current decks.
+
+## Current manuscript figure scopes
+
+The user's 2026-10-13 definitions supersede old figure numbers inside the deck and earlier plans:
+Figure 1 overview/corpus statistics; Figure 2 architecture/gene-context ablations on SLE pseudobulk;
+Figure 3 healthy measured bulk–sc paired mismatch, full FT, LoRA ranks and support budgets;
+Figure 4 disease measured-paired quantitative integrity comparisons;
+Figure 5 paired rare-population qualitative recovery; Figure 6 paired adaptation followed by
+bulk-only published-finding reproduction (TODO).
+The authoritative scope and proposed 131-page mapping are in
+`presentations/2026-09-30--paper-summary--r00/data/figure_assignment_20261013/`.
+Read `AUTHOR_SCOPE.md` and `ASSIGNMENT.md` before figure work. Preserve the distinction
+between total-five pairs and five training plus extra validation; rank16 is not universally
+best in existing metrics, and disease comparisons have cohort/endpoint exceptions.
+Existing rare-state quantitative slides do not substitute for paired qualitative panels.
+The scope is now applied in the actual reordered PPTX/PDF, not only in planning documents.
+Use `tools/build_figure1_focused_master_20261015.py` in the presentation directory. It consumes
+the verified 155-page spec and the source-bound Figure 1 focus bundles; older builders are preserved historical entry points.
+Figure 2 now includes corpus-validation curves and cost metadata alongside SLE pseudobulk
+benchmarks; these are not paired real-bulk performance panels. Figure 1's recorded PCA
+uses an existing ctx2048 E12 illustration, with endpoint colours revealed by denoising progress,
+not per-step confidence. Display real R2 as a real-cell repeat from the same donors, not new donors.
+Healthy pair-budget panels cover all six types with both Pearson correlation and MAE, including adverse results.
+They use historical n+7 validation access, not total-n. Do not omit NK/DC/Other to make a five-pair claim appear stronger.
+
+The latest strengthening proposal is `data/figure_assignment_20261013/STRENGTHENING_PROPOSAL.md` under the deck.
+Corpus 41,395,184 vs train 33,141,380 is a split difference, not K exclusion; all training cells are covered after top-up.
+Healthy 88.51% is a source-cell share. The remainder includes ACPA+ at-risk and is labelled Disease / at-risk.
+Historical LoRA all means 78–79 training +7 validation, keeping21–22 test/fold. Nested random training subsets and
+outer/inner splits replay exactly. The legacy total-five102-test/draw design differs from strict healthy outer-fold tests.
+Figure5 original-paper targets require author identity: Ye Mono_c06 pediatric pulmonary and Imperial wave2
+steroid-associated monocytes. Ye observed-marker co-detection is a diagnostic, not reproduced cluster identity;
+fine labels were not found in133 audited canonical obs schemas. Existing generic rare-state rules do not substitute
+for those source-defined populations. Giroux/Ji program results are not automatically rare-population evidence.
+
+Figure 1 main corpus panel now uses exact train metadata: 33,141,380 cells, 5,232 samples, 2,572 corrected donor identities; Healthy 88.4241%. The 88.51% healthy share applies to the full source pool only. Training cell-type proportions were recomputed across 432 metadata shards, not scaled from full-source proportions. Use `data/figure1_revision_20261014/train_corpus/result.json`. Figure 1a uses organic vector glyphs; cells/model nodes are schematic, and K is distinct from generated-cell count.
+
+Figure 1a now explicitly shows noisy gene-space state -> iterative PBISC denoising -> sparse expression with exact zeros; those profiles are schematic, not recorded data. Figure 1b main proportions use grouped small categories with source membership in `data/figure1_focus_20261015/corpus/display_aggregations.csv`; Unlabelled and ACPA+ at-risk remain separate. Complete training counts are preserved on support page9, full-source composition page10. Do not revert the denoiser to a generic molecular-network icon or drop the detail panel.

paper-releases/pbisc-diffusion/updates/20261015T100456Z-figure1-diffusion-readable-corpus/diffusion/dataset/server_1/model8/ara/trace/exploration_tree.yaml

@@ -0,0 +1,69 @@
+tree:
+- id: N01
+  type: decision
+  provenance: user
+  choice: Ground Figure5 targets in each paired cohort original paper.
+  evidence:
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/data/strengthening_20261013/rare_literature/result.json
+- id: N02
+  type: experiment
+  provenance: ai-executed
+  result: Observed Ye marker co-detection is present but does not reproduce pulmonary
+    specificity; no author fine labels found in133schemas.
+  evidence:
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/data/strengthening_20261013/rare_literature/result.json
+- id: N03
+  type: experiment
+  provenance: ai-executed
+  result: Launched2checkpointparents x5LoRA draws for representative paired Sepsis;
+    firsttrunkdraw completed3000steps and sampling started. Finalcomparison pending.
+  evidence:
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/trunk4x_lora_sepsis_20261013/PROTOCOL.json
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/trunk4x_lora_sepsis_20261013/LAUNCH_RECEIPT.json
+- id: N04
+  type: decision
+  provenance: user
+  choice: Pretrain trunkx4 with the exact current-paper f01 6k gene context; prefer
+    server2, fallback server1.
+  evidence:
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/f01_trunk4x_complete_20261013/PLAN.md
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/f01_trunk4x_complete_20261013/CONTRACT.json
+- id: N05
+  type: experiment
+  provenance: ai-executed
+  result: 128-stepGPU profile passed, actual genes/positions match f01, full16parttraining
+    launched onserver1 with NAS-backed output.100adaptations scheduled; no outcome
+    superiority claimed.
+  evidence:
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/f01_trunk4x_complete_20261013/LAUNCH_RECEIPT.json
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/f01_trunk4x_complete_20261013/PROFILE_GENE_PROOF.json
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/f01_trunk4x_complete_20261013/MATRIX_VERIFIED.json
+- id: N06
+  type: decision
+  provenance: user
+  timestamp: '2026-10-14'
+  choice: Redraw Figure1a in supplied organic schematic style and add exact33.14M
+    training-corpus main panel.
+  evidence:
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/data/figure1_revision_20261014/overview/result.json
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/data/figure1_revision_20261014/train_corpus/result.json
+- id: N07
+  type: experiment
+  provenance: ai-executed
+  timestamp: '2026-10-14'
+  result: Metadata-only aggregation across432shards verified33,141,380trainingcells,5232samples,2572corrected
+    donors,Healthy88.4241%;155-page deck rendered and verified. No model training
+    or inference.
+  evidence:
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/data/figure1_revision_20261014/train_corpus/VERIFICATION.json
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/data/figure1_revision_20261014/qa/VERIFICATION.json
+- id: N08
+  type: decision
+  provenance: user
+  timestamp: '2026-10-15'
+  choice: 'Execute Figure1 review: explicit diffusion mechanism and readable main
+    corpus proportions with complete support details.'
+  evidence:
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/data/figure1_focus_20261015/overview/result.json
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/data/figure1_focus_20261015/corpus/result.json
+  - /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/data/figure1_focus_20261015/qa/VERIFICATION.json

paper-releases/pbisc-diffusion/updates/20261015T100456Z-figure1-diffusion-readable-corpus/diffusion/dataset/server_1/model8/ara/trace/pm_reasoning_log.yaml

@@ -0,0 +1,23 @@
+entries:
+- date: '2026-10-13'
+  provenance: ai-executed
+  note: Recorded source/audit actions without promoting universal superiority or rare-population
+    identity claims. No existing paper-specific ARA found; parent artifact covers
+    broader models and was left unchanged.
+- date: '2026-10-13'
+  provenance: ai-executed
+  note: Checkpointpanels differ; addmatchedold-panelx1control rather than falselyattributeallgainstowidth.
+    Keep superiority hypothesis staged until all prespecified draws finish.
+- date: '2026-10-13'
+  provenance: ai-executed
+  note: Architecture comparison now fixes all6000genes and source stream. Resourcehost
+    follows observedserver2busy→server1fallback. Profile success does not establish
+    E16performance; all scientific claims remain pending.
+- date: '2026-10-14'
+  provenance: ai-executed
+  note: Recorded requested figure changes and metadata arithmetic. No model-performance
+    or rare-population claim promoted; prior trace preserved.
+- date: '2026-10-15'
+  provenance: ai-executed
+  note: Visual revision only; no new efficacy claim. Saved counts preserved, small-category
+    groups auditable and full detail retained.

paper-releases/pbisc-diffusion/updates/20261015T100456Z-figure1-diffusion-readable-corpus/diffusion/dataset/server_1/model8/ara/trace/sessions/2026-10-15_001.yaml

@@ -0,0 +1,16 @@
+id: 2026-10-15_001
+date: '2026-10-15'
+summary: Implemented Figure1 suggestions in156-slide master; all numerical source
+  counts preserved.
+ai_actions:
+- provenance: ai-executed
+  action: Made diffusion mechanism explicit, enlarged corpus legend, retained full
+    training detail as support and exported2-page preview.
+- provenance: ai-executed
+  action: Verified156PPTX/PDFpages and source counts without new model work or metadata
+    rescan.
+events:
+- N08
+claims_touched: []
+open_threads:
+- Existing Figure2–6 work remains outside this Figure1 task.