Datasets:

junidude14
/

PBISC-data

paper-releases/pbisc-diffusion/updates/20260930T065633Z-summary-deck-scsemi-final/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/sec03_server2_snapshots/f01_3k_1x_seed20260627/RUN_STATUS.json ADDED Viewed

	@@ -0,0 +1,38 @@

+{
+  "best_validation_loss": 0.10918324212700364,
+  "completed_at": "2026-09-24T19:57:45.600148+00:00",
+  "condition_dim": 9006,
+  "corpus_fingerprint": "eeb609250457248225432b6df94573b26a8cad8ac743670b5ca43134fea7e007",
+  "device": "cuda",
+  "epochs_completed": 16,
+  "generated_zero_fraction_preview": 0.905390739440918,
+  "global_step": 764249,
+  "memory": {
+    "cuda_allocated_gib": 0.5265121459960938,
+    "cuda_free_gib": 9.902095794677734,
+    "cuda_peak_allocated_gib": 0.7942924499511719,
+    "cuda_reserved_gib": 0.8671875,
+    "cuda_total_gib": 121.68955612182617,
+    "system_available_gib": 104.14525985717773
+  },
+  "memory_policy": {
+    "cuda_allocator_fraction": 0.06574105662766178,
+    "cuda_allocator_limit_gib": 8.0,
+    "cuda_available_at_start_gib": 41.15022659301758,
+    "cuda_device_index": 0,
+    "cuda_memory_floor_gib": 0.0,
+    "cuda_temperature_at_start_c": 47.0,
+    "cuda_utilization_at_start_percent": 0.0,
+    "cuda_visible_total_gib": 121.68955612182617,
+    "system_available_at_start_gib": 116.39532852172852,
+    "system_memory_floor_gib": 32.0
+  },
+  "n_genes": 3000,
+  "parameters": 31139184,
+  "real_zero_fraction_preview": 0.8917739391326904,
+  "schema_version": "pbisc_sdd.status.v1",
+  "status": "COMPLETED",
+  "support_value_disagreement_fraction_preview": 1.372661017740029e-06,
+  "training_contract": "pbisc_sdd.ladder_training.v1",
+  "wall_seconds": 38160.57213782612
+}

	@@ -0,0 +1,85 @@

+{
+  "data": {
+    "coverage_audit_sha256": "3904c5b4026f3a5358d6553948f0d4456c80494eab941ab2f7c2975f459d24c5",
+    "coverage_audit_summary": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/full_recipe_allcells_6k_broad_20260909/coverage_v1/summary.json",
+    "gene_positions_path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/artifacts/panel_ladder_floor01_1m_20260924_r2/tiers/3k/gene_positions.npy",
+    "gene_positions_sha256": "8ecfbc9c1bb2ee61eb6d9c7c2aabeedaceb8a4f478f9a93e707fd826edcc1121",
+    "gene_selection": "explicit_train_only_hvg_manifest",
+    "gene_selection_manifest": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/artifacts/panel_ladder_floor01_1m_20260924_r2/tiers/3k/legacy_selection_manifest.v1.json",
+    "gene_selection_manifest_sha256": "a58c31d6fefaf30f95eef32e95113d22068d41d36d9c106fccb10ed79d786d4a",
+    "include_availability_in_condition": true,
+    "include_bulk_measurement_in_condition": true,
+    "n_genes": 3000,
+    "priming_policy": "all-train-first-then-validation-v1",
+    "profile_worst_k": false,
+    "progress": false,
+    "test_cells_per_block": 64,
+    "test_recipes": 192,
+    "topup_manifest": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/full_recipe_allcells_6k_broad_20260909/topup_v1/manifest.json",
+    "topup_manifest_sha256": "df300a6f7f758c865209c6640e51abcbec10bab9726ee48c3bc3434936316da1",
+    "train_cells_per_block": 4000,
+    "train_recipes": 8192,
+    "validation_cells_per_block": 64,
+    "validation_recipes": 192
+  },
+  "diffusion": {
+    "center_values": true,
+    "clip_denoised": true,
+    "cosine_offset": 0.008,
+    "log_snr_clip": 20.0,
+    "max_log1p_value": 8.0,
+    "self_condition_probability": 0.5,
+    "support_loss_weight": 1.0,
+    "support_scale": 1.0,
+    "value_loss_mode": "all_measured",
+    "value_loss_weight": 1.0
+  },
+  "experiment_id": "f01_3k_1x_seed20260627",
+  "model": {
+    "context_dim": 512,
+    "dropout": 0.0,
+    "hidden_dims": [
+      1024,
+      1024,
+      512,
+      256
+    ],
+    "time_embedding_dim": 128
+  },
+  "paths": {
+    "conditioner_cache": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_7_no_annotation_sh_hwang/runs/E2_full_no_annotation_4epoch_seed20260627/conditioner_cache.pt",
+    "model7_root": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_7_no_annotation_sh_hwang",
+    "output_dir": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/runs/f01_3k_1x_seed20260627"
+  },
+  "sampling": {
+    "active_logit_threshold": 0.0,
+    "eta": 0.0,
+    "preview_cells": 256,
+    "steps": 50
+  },
+  "schema_version": "pbisc_sdd.train.v1",
+  "training": {
+    "amp": "bf16",
+    "batch_size": 256,
+    "device": "cuda",
+    "ema_decay": 0.999,
+    "ema_update_every": 1,
+    "epochs": 16,
+    "grad_clip": 5.0,
+    "gradient_accumulation_steps": 1,
+    "heartbeat_every_steps": 100,
+    "lr": 0.0002,
+    "max_cuda_memory_gib": 8.0,
+    "max_cuda_temperature_c_at_start": 75.0,
+    "max_cuda_utilization_percent_at_start": 10.0,
+    "max_train_steps": 0,
+    "max_validation_steps": 0,
+    "min_cuda_memory_available_gib": 0.0,
+    "min_system_memory_available_gib": 32.0,
+    "num_workers": 4,
+    "pin_memory": true,
+    "prefetch_factor": 32,
+    "seed": 20260627,
+    "weight_decay": 0.0
+  }
+}

	@@ -0,0 +1,629 @@

+{
+  "epochs": [
+    {
+      "epoch": 1,
+      "global_step": 47765,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 7.643585205078125,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 108.10719299316406
+      },
+      "train": {
+        "batches": 47765,
+        "loss": 0.15872636194051806,
+        "optimizer_steps": 47765,
+        "support_accuracy": 0.9406105036250081,
+        "support_elements": 35810984447,
+        "support_f1": 0.6577419887886128,
+        "support_loss": 0.1543494444309125,
+        "support_precision": 0.8119276667279492,
+        "support_recall": 0.5527704603335687,
+        "value_elements": 35810984447,
+        "value_loss": 0.004376917509605553
+      },
+      "train_seconds": 2337.69444428687,
+      "validation": {
+        "batches": 48,
+        "loss": 0.12453482428235711,
+        "support_accuracy": 0.9555090986855006,
+        "support_elements": 35576128,
+        "support_f1": 0.6954277232211409,
+        "support_loss": 0.12176107409202457,
+        "support_precision": 0.84443813054903,
+        "support_recall": 0.5911183761806329,
+        "value_elements": 35576128,
+        "value_loss": 0.0027737501903325416
+      },
+      "validation_seconds": 25.352466464973986
+    },
+    {
+      "epoch": 2,
+      "global_step": 95530,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.45062255859375,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 105.53097534179688
+      },
+      "train": {
+        "batches": 47765,
+        "loss": 0.1307791093537549,
+        "optimizer_steps": 47765,
+        "support_accuracy": 0.9521386259374904,
+        "support_elements": 35804889863,
+        "support_f1": 0.7372591367729396,
+        "support_loss": 0.12697468315336574,
+        "support_precision": 0.8551561766965932,
+        "support_recall": 0.6479313709349072,
+        "value_elements": 35804889863,
+        "value_loss": 0.0038044262003891485
+      },
+      "train_seconds": 2339.6700600238983,
+      "validation": {
+        "batches": 48,
+        "loss": 0.11713109341609906,
+        "support_accuracy": 0.95820079127217,
+        "support_elements": 35576128,
+        "support_f1": 0.718410476340522,
+        "support_loss": 0.11443468107129585,
+        "support_precision": 0.8529442842966258,
+        "support_recall": 0.6205343838824499,
+        "value_elements": 35576128,
+        "value_loss": 0.002696412344803206
+      },
+      "validation_seconds": 25.196826808154583
+    },
+    {
+      "epoch": 3,
+      "global_step": 143293,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.342987060546875,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 105.38168334960938
+      },
+      "train": {
+        "batches": 47763,
+        "loss": 0.12535752799624114,
+        "optimizer_steps": 47763,
+        "support_accuracy": 0.9542227060430495,
+        "support_elements": 35825485437,
+        "support_f1": 0.7484228484933073,
+        "support_loss": 0.1216207461174,
+        "support_precision": 0.8615509879158977,
+        "support_recall": 0.6615556098538575,
+        "value_elements": 35825485437,
+        "value_loss": 0.0037367818788411563
+      },
+      "train_seconds": 2240.2958074430935,
+      "validation": {
+        "batches": 48,
+        "loss": 0.11482954749147845,
+        "support_accuracy": 0.9590090860927867,
+        "support_elements": 35576128,
+        "support_f1": 0.7244411909871001,
+        "support_loss": 0.11214786314815807,
+        "support_precision": 0.8576051345580074,
+        "support_recall": 0.627072946452025,
+        "value_elements": 35576128,
+        "value_loss": 0.002681684343320382
+      },
+      "validation_seconds": 25.07160533592105
+    },
+    {
+      "epoch": 4,
+      "global_step": 191057,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.414260864257812,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 105.33919906616211
+      },
+      "train": {
+        "batches": 47764,
+        "loss": 0.12404838276044054,
+        "optimizer_steps": 47764,
+        "support_accuracy": 0.9546456095837967,
+        "support_elements": 35800998362,
+        "support_f1": 0.7530584386509992,
+        "support_loss": 0.12030716471048286,
+        "support_precision": 0.8637059444328964,
+        "support_recall": 0.6675411780240939,
+        "value_elements": 35800998362,
+        "value_loss": 0.003741218049957676
+      },
+      "train_seconds": 2160.8869622820057,
+      "validation": {
+        "batches": 48,
+        "loss": 0.11347530155151482,
+        "support_accuracy": 0.9595155774119095,
+        "support_elements": 35576128,
+        "support_f1": 0.72836553558486,
+        "support_loss": 0.11079471243769291,
+        "support_precision": 0.8600057541704375,
+        "support_recall": 0.6316755868039324,
+        "value_elements": 35576128,
+        "value_loss": 0.0026805891138219134
+      },
+      "validation_seconds": 24.6442953790538
+    },
+    {
+      "epoch": 5,
+      "global_step": 238822,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.302959442138672,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 104.98669815063477
+      },
+      "train": {
+        "batches": 47765,
+        "loss": 0.12212307805809793,
+        "optimizer_steps": 47765,
+        "support_accuracy": 0.9553918892171148,
+        "support_elements": 35814749761,
+        "support_f1": 0.7564076482745801,
+        "support_loss": 0.11841846108455674,
+        "support_precision": 0.8656410709686222,
+        "support_recall": 0.6716531522752476,
+        "value_elements": 35814749761,
+        "value_loss": 0.0037046169735411873
+      },
+      "train_seconds": 2351.858669494977,
+      "validation": {
+        "batches": 48,
+        "loss": 0.11256631475630774,
+        "support_accuracy": 0.9598673582465186,
+        "support_elements": 35576128,
+        "support_f1": 0.730360666338312,
+        "support_loss": 0.10988049670329117,
+        "support_precision": 0.8639566570024221,
+        "support_recall": 0.6325480275870552,
+        "value_elements": 35576128,
+        "value_loss": 0.002685818053016571
+      },
+      "validation_seconds": 24.994933484122157
+    },
+    {
+      "epoch": 6,
+      "global_step": 286586,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.238288879394531,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 105.05133056640625
+      },
+      "train": {
+        "batches": 47764,
+        "loss": 0.12110101012496519,
+        "optimizer_steps": 47764,
+        "support_accuracy": 0.9557813947243361,
+        "support_elements": 35801824710,
+        "support_f1": 0.7587287379424513,
+        "support_loss": 0.11741043214181061,
+        "support_precision": 0.8670368113268556,
+        "support_recall": 0.674475004605597,
+        "value_elements": 35801824710,
+        "value_loss": 0.0036905779831545695
+      },
+      "train_seconds": 2322.0651968098246,
+      "validation": {
+        "batches": 48,
+        "loss": 0.11190816275074006,
+        "support_accuracy": 0.9601140124074211,
+        "support_elements": 35576128,
+        "support_f1": 0.7323429379391077,
+        "support_loss": 0.10923583307819172,
+        "support_precision": 0.8648685855425831,
+        "support_recall": 0.6350351544239814,
+        "value_elements": 35576128,
+        "value_loss": 0.0026723296725483345
+      },
+      "validation_seconds": 24.75596708804369
+    },
+    {
+      "epoch": 7,
+      "global_step": 334356,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.258590698242188,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 105.0426139831543
+      },
+      "train": {
+        "batches": 47770,
+        "loss": 0.1196841914425701,
+        "optimizer_steps": 47770,
+        "support_accuracy": 0.9563717889141377,
+        "support_elements": 35846916412,
+        "support_f1": 0.7608380412984318,
+        "support_loss": 0.11601985200620879,
+        "support_precision": 0.8684544869313746,
+        "support_recall": 0.6769520418539478,
+        "value_elements": 35846916412,
+        "value_loss": 0.0036643394363613134
+      },
+      "train_seconds": 2371.29087424092,
+      "validation": {
+        "batches": 48,
+        "loss": 0.11139553138693431,
+        "support_accuracy": 0.9602785609496346,
+        "support_elements": 35576128,
+        "support_f1": 0.7338034496756687,
+        "support_loss": 0.10872749495105946,
+        "support_precision": 0.8650186483478515,
+        "support_recall": 0.6371532849341874,
+        "value_elements": 35576128,
+        "value_loss": 0.0026680364358748526
+      },
+      "validation_seconds": 24.968876827973872
+    },
+    {
+      "epoch": 8,
+      "global_step": 382120,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.344623565673828,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 104.95572280883789
+      },
+      "train": {
+        "batches": 47764,
+        "loss": 0.11978765599122215,
+        "optimizer_steps": 47764,
+        "support_accuracy": 0.9562911198131961,
+        "support_elements": 35822327438,
+        "support_f1": 0.7612031561597951,
+        "support_loss": 0.11613016646580683,
+        "support_precision": 0.8687406931929696,
+        "support_recall": 0.6773562464767412,
+        "value_elements": 35822327438,
+        "value_loss": 0.0036574895254153266
+      },
+      "train_seconds": 2367.2427881869953,
+      "validation": {
+        "batches": 48,
+        "loss": 0.1110197367153232,
+        "support_accuracy": 0.9604235177026572,
+        "support_elements": 35576128,
+        "support_f1": 0.7346631790410788,
+        "support_loss": 0.10835318062687914,
+        "support_precision": 0.8665276097729256,
+        "support_recall": 0.6376312136871467,
+        "value_elements": 35576128,
+        "value_loss": 0.0026665560884440633
+      },
+      "validation_seconds": 24.488298936048523
+    },
+    {
+      "epoch": 9,
+      "global_step": 429884,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.247207641601562,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 104.8751449584961
+      },
+      "train": {
+        "batches": 47764,
+        "loss": 0.11984249623687131,
+        "optimizer_steps": 47764,
+        "support_accuracy": 0.9562583574166901,
+        "support_elements": 35816951410,
+        "support_f1": 0.7623476662182693,
+        "support_loss": 0.11617504307640739,
+        "support_precision": 0.8691939158577806,
+        "support_recall": 0.6788941445946431,
+        "value_elements": 35816951410,
+        "value_loss": 0.003667453160463924
+      },
+      "train_seconds": 2465.179816304939,
+      "validation": {
+        "batches": 48,
+        "loss": 0.11073773451641142,
+        "support_accuracy": 0.9605437668764852,
+        "support_elements": 35576128,
+        "support_f1": 0.7358579470120166,
+        "support_loss": 0.10805509411318986,
+        "support_precision": 0.8662077000945844,
+        "support_recall": 0.6396076993308016,
+        "value_elements": 35576128,
+        "value_loss": 0.0026826404032215556
+      },
+      "validation_seconds": 24.710099382093176
+    },
+    {
+      "epoch": 10,
+      "global_step": 477660,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.257404327392578,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 104.80999374389648
+      },
+      "train": {
+        "batches": 47776,
+        "loss": 0.11878660788576692,
+        "optimizer_steps": 47776,
+        "support_accuracy": 0.956676335027382,
+        "support_elements": 35827798548,
+        "support_f1": 0.7630585038384097,
+        "support_loss": 0.11515039539149607,
+        "support_precision": 0.8697718444438898,
+        "support_recall": 0.679669095305909,
+        "value_elements": 35827798548,
+        "value_loss": 0.0036362124942708493
+      },
+      "train_seconds": 2403.702841494931,
+      "validation": {
+        "batches": 48,
+        "loss": 0.11041147969322394,
+        "support_accuracy": 0.9606684853393826,
+        "support_elements": 35576128,
+        "support_f1": 0.7364230152324136,
+        "support_loss": 0.1077208346150799,
+        "support_precision": 0.8680777723105564,
+        "support_recall": 0.6394434828804351,
+        "value_elements": 35576128,
+        "value_loss": 0.0026906450781440396
+      },
+      "validation_seconds": 24.384688103105873
+    },
+    {
+      "epoch": 11,
+      "global_step": 525426,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.345634460449219,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 104.67244338989258
+      },
+      "train": {
+        "batches": 47766,
+        "loss": 0.11870569791172018,
+        "optimizer_steps": 47766,
+        "support_accuracy": 0.9566804391284536,
+        "support_elements": 35820213866,
+        "support_f1": 0.7641638705962965,
+        "support_loss": 0.11506706161240808,
+        "support_precision": 0.8701742400909707,
+        "support_recall": 0.6811782411560012,
+        "value_elements": 35820213866,
+        "value_loss": 0.0036386362993121096
+      },
+      "train_seconds": 2396.3314447500743,
+      "validation": {
+        "batches": 48,
+        "loss": 0.11010085496292613,
+        "support_accuracy": 0.9607649826310497,
+        "support_elements": 35576128,
+        "support_f1": 0.7375904257719089,
+        "support_loss": 0.10742881937469936,
+        "support_precision": 0.8671224781078344,
+        "support_recall": 0.6417281197118295,
+        "value_elements": 35576128,
+        "value_loss": 0.0026720355882267574
+      },
+      "validation_seconds": 25.07389808516018
+    },
+    {
+      "epoch": 12,
+      "global_step": 573189,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.354778289794922,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 104.77531814575195
+      },
+      "train": {
+        "batches": 47763,
+        "loss": 0.11930658929875551,
+        "optimizer_steps": 47763,
+        "support_accuracy": 0.9563877668703732,
+        "support_elements": 35781902829,
+        "support_f1": 0.7645005335405294,
+        "support_loss": 0.11564748067471457,
+        "support_precision": 0.8700587406575838,
+        "support_recall": 0.6817843549781916,
+        "value_elements": 35781902829,
+        "value_loss": 0.0036591086240409386
+      },
+      "train_seconds": 2320.4146038671024,
+      "validation": {
+        "batches": 48,
+        "loss": 0.10981549890805548,
+        "support_accuracy": 0.9608713179804166,
+        "support_elements": 35576128,
+        "support_f1": 0.7382152411131118,
+        "support_loss": 0.10714198869125545,
+        "support_precision": 0.8682442236165522,
+        "support_recall": 0.6420598238565939,
+        "value_elements": 35576128,
+        "value_loss": 0.0026735102168000194
+      },
+      "validation_seconds": 24.41797095607035
+    },
+    {
+      "epoch": 13,
+      "global_step": 620952,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.367931365966797,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 104.60245132446289
+      },
+      "train": {
+        "batches": 47763,
+        "loss": 0.11787352866482213,
+        "optimizer_steps": 47763,
+        "support_accuracy": 0.9569958358956324,
+        "support_elements": 35824211122,
+        "support_f1": 0.7655213548766521,
+        "support_loss": 0.11425957649437082,
+        "support_precision": 0.8710884325834454,
+        "support_recall": 0.6827758509553591,
+        "value_elements": 35824211122,
+        "value_loss": 0.0036139521704513103
+      },
+      "train_seconds": 2445.1615354609676,
+      "validation": {
+        "batches": 48,
+        "loss": 0.10962550971520928,
+        "support_accuracy": 0.9609279851927675,
+        "support_elements": 35576128,
+        "support_f1": 0.7385439177172798,
+        "support_loss": 0.10694881038920052,
+        "support_precision": 0.868857537624615,
+        "support_recall": 0.6422217504361386,
+        "value_elements": 35576128,
+        "value_loss": 0.002676699326008759
+      },
+      "validation_seconds": 24.752229939913377
+    },
+    {
+      "epoch": 14,
+      "global_step": 668718,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.263614654541016,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 104.62947082519531
+      },
+      "train": {
+        "batches": 47766,
+        "loss": 0.11775740981217442,
+        "optimizer_steps": 47766,
+        "support_accuracy": 0.9570225777748669,
+        "support_elements": 35817227728,
+        "support_f1": 0.7656326442688652,
+        "support_loss": 0.1141535155209273,
+        "support_precision": 0.8711019805015411,
+        "support_recall": 0.6829446056122039,
+        "value_elements": 35817227728,
+        "value_loss": 0.0036038942912471152
+      },
+      "train_seconds": 2433.8989064369816,
+      "validation": {
+        "batches": 48,
+        "loss": 0.10950227480989054,
+        "support_accuracy": 0.9609585112803731,
+        "support_elements": 35576128,
+        "support_f1": 0.7387241067500075,
+        "support_loss": 0.10681062159192312,
+        "support_precision": 0.8691779087659158,
+        "support_recall": 0.6423192335082686,
+        "value_elements": 35576128,
+        "value_loss": 0.0026916532179674184
+      },
+      "validation_seconds": 24.526882331818342
+    },
+    {
+      "epoch": 15,
+      "global_step": 716481,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.026630401611328,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 104.37038040161133
+      },
+      "train": {
+        "batches": 47763,
+        "loss": 0.11786515801571865,
+        "optimizer_steps": 47763,
+        "support_accuracy": 0.9569631604313669,
+        "support_elements": 35810937477,
+        "support_f1": 0.7662934662518386,
+        "support_loss": 0.11425504138967,
+        "support_precision": 0.8713796835227239,
+        "support_recall": 0.6838257818991608,
+        "value_elements": 35810937477,
+        "value_loss": 0.0036101166260486444
+      },
+      "train_seconds": 2327.106154473964,
+      "validation": {
+        "batches": 48,
+        "loss": 0.10930036411464983,
+        "support_accuracy": 0.9610268998357551,
+        "support_elements": 35576128,
+        "support_f1": 0.7395358762978007,
+        "support_loss": 0.1066296999156392,
+        "support_precision": 0.8685374145907748,
+        "support_recall": 0.6438992443753412,
+        "value_elements": 35576128,
+        "value_loss": 0.002670664199010634
+      },
+      "validation_seconds": 24.790227458113804
+    },
+    {
+      "epoch": 16,
+      "global_step": 764249,
+      "memory": {
+        "cuda_allocated_gib": 0.5265121459960938,
+        "cuda_free_gib": 10.002067565917969,
+        "cuda_peak_allocated_gib": 0.7942924499511719,
+        "cuda_reserved_gib": 0.8671875,
+        "cuda_total_gib": 121.68955612182617,
+        "system_available_gib": 104.4130859375
+      },
+      "train": {
+        "batches": 47768,
+        "loss": 0.11740522085380325,
+        "optimizer_steps": 47768,
+        "support_accuracy": 0.9571497991981235,
+        "support_elements": 35824604699,
+        "support_f1": 0.767359792144282,
+        "support_loss": 0.11378709298971332,
+        "support_precision": 0.8718234087036217,
+        "support_recall": 0.6852515831695458,
+        "value_elements": 35824604699,
+        "value_loss": 0.003618127864089933
+      },
+      "train_seconds": 2372.1610324829817,
+      "validation": {
+        "batches": 48,
+        "loss": 0.10918324212700364,
+        "support_accuracy": 0.9610783669318932,
+        "support_elements": 35576128,
+        "support_f1": 0.7394713229194438,
+        "support_loss": 0.10650228794077275,
+        "support_precision": 0.8703052506080743,
+        "support_recall": 0.6428334730699742,
+        "value_elements": 35576128,
+        "value_loss": 0.0026809541862308863
+      },
+      "validation_seconds": 24.694192191120237
+    }
+  ],
+  "training_contract": "pbisc_sdd.ladder_training.v1"
+}

paper-releases/pbisc-diffusion/updates/20260930T065633Z-summary-deck-scsemi-final/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/slides_spec.json ADDED Viewed

	@@ -0,0 +1,636 @@

+{
+  "deck": {
+    "output_pptx": "PBISC-Diffusion_paper_summary_20260930.pptx",
+    "paths": "figure paths are relative to the deck folder",
+    "note": "footnote = on-slide text (max 2 lines, may be a shortened copy); footnote_full = section author's full text, put in speaker notes when it differs. Section 9 text comes from data/sec09_slide_text.json; refresh with tools/sync_sec09_text.py after rerunning tools/sec09_paired_validation.py.",
+    "moved": "The ImMicro external-reference slide (section 4 spec) is placed at the end of section 5, after refcal is introduced, as its author asked."
+  },
+  "slides": [
+    {
+      "kind": "title",
+      "title": "PBISC-Diffusion: bulk-to-single-cell generation",
+      "subtitle": "Paper summary · 2026-09-30"
+    },
+    {
+      "kind": "agenda",
+      "title": "Outline",
+      "items": [
+        "Model overview: diffusion from bulk to cells",
+        "Backbone: SDD, scDiffusion, bulk conditioning",
+        "Architecture ablation: width, context, gene panel",
+        "SLE benchmark vs scSemiProfiler",
+        "Real bulk ≠ pseudobulk",
+        "Fine-tuning on paired bulk–sc data",
+        "LoRA: compute and ablations (K diversification)",
+        "How many bulk–sc pairs are needed",
+        "Paired-cohort validation vs scSemiProfiler",
+        "Next: cohorts with a paired subset (planned)"
+      ]
+    },
+    {
+      "kind": "content",
+      "section": "1",
+      "section_label": "1 Model overview",
+      "title": "PBISC-Diffusion: bulk in, cells out",
+      "layout": "full",
+      "figures": [
+        "figures/sec01_slide_a.png"
+      ],
+      "keywords": [
+        "Train pair: K-cell pseudobulk → each member cell"
+      ],
+      "footnote": "Fig. 1b redraw. Final model corpus: 33.1M train cells; K = 1–4,000 cells per pseudobulk.",
+      "footnote_full": "Top: deck redraw of manuscript Fig. 1b (larger labels; cells differ by x_T and the Bernoulli support draw, Methods 'Support read-out'). Bottom: corpus of final run f01_6k_1x_seed20260627 (index fingerprint matches). Cells per dataset = base + overlay rows (Model 7 e2_full_source/e2_full_index manifests); 33.1M cells in the donor-level 80/10/10 train split. Registered K grid = 8,192 recipes/epoch. Plus 3,658 top-up recipes over 16 epochs (196-283 per epoch, K 1-4,000), so every train cell is a target once (ALLCELL_DATA_CONTRACT.json coverage_after_topup).",
+      "sources": "tools/sec01_overview.py; data/sec01_model_numbers.csv, data/sec01_corpus_datasets.csv, data/sec01_recipe_k.csv"
+    },
+    {
+      "kind": "content",
+      "section": "1",
+      "section_label": "1 Model overview",
+      "title": "Denoiser: joint value + support diffusion",
+      "layout": "diagram",
+      "figures": [
+        "figures/sec01_fig1a_denoiser.png"
+      ],
+      "keywords": [
+        "6,000 genes · 54.2M params · Bernoulli support draw"
+      ],
+      "footnote": "Fig. 1a redraw. Final model f01 6k 1x, epoch 16, EMA; 50 DDIM steps.",
+      "footnote_full": "Deck redraw of manuscript Fig. 1a (larger labels, legend dropped). Purple dashed = soft self-conditioning feedback; gold dashed = context to all blocks. Final model f01_6k_1x_seed20260627, checkpoint_epoch_016, EMA; 7 blocks 1024-1024-512-256-512-1024-1024 (effective_config.json, RUN_STATUS.json). Sampling: 50 DDIM steps, η = 0, drawn support read-out, keep gene if u < σ(logit) (methods.tex l.58, l.99). 16 epochs = 16 disjoint 8,192-recipe slots = one pass over the 131,072-recipe pool (manuscript: part 16).",
+      "sources": "tools/sec01_overview.py; data/sec01_model_numbers.csv, data/sec01_fig1_label_pt.csv"
+    },
+    {
+      "kind": "content",
+      "section": "2",
+      "section_label": "2 Backbone",
+      "title": "Backbone: SDD → PBISC-Diffusion",
+      "layout": "diagram",
+      "figures": [
+        "figures/sec02_backbone_compare.png"
+      ],
+      "keywords": [
+        "Kept from SDD: value + bit, skip-MLP",
+        "Changed: bulk condition, masked loss, Bernoulli read-out"
+      ],
+      "footnote": "SDD: Ostheimer et al., arXiv 2502.02448. scDiffusion: Luo et al., Bioinformatics 2024.",
+      "sources": "tools/sec02_backbone.py; data/sec02_backbone_compare.csv, data/sec02_slide_text.csv, data/refs/",
+      "footnote_full": "Luo et al., Bioinformatics 2024; Ostheimer et al., arXiv 2502.02448 (code 1d5423c). PBISC: final model f01 6k 1x, part 16. σ = sigmoid; Bernoulli is the final read-out; the logit > 0 cut is also reported."
+    },
+    {
+      "kind": "content",
+      "section": "2",
+      "section_label": "2 Backbone",
+      "title": "Bulk conditioning",
+      "layout": "diagram",
+      "figures": [
+        "figures/sec02_bulk_conditioning.png"
+      ],
+      "keywords": [
+        "Same bulk condition for every generated cell",
+        "Final: no cell types, proportions, sum constraint",
+        "18,006 → 512 context into every block"
+      ],
+      "footnote": "Final model f01 6k 1x. Bar heights schematic.",
+      "sources": "tools/sec02_backbone.py; data/sec02_bulk_conditioning.csv, data/sec02_slide_text.csv",
+      "footnote_full": "Final model f01 6k 1x, part 16. Bar heights schematic."
+    },
+    {
+      "kind": "content",
+      "section": "3",
+      "section_label": "3 Architecture ablation",
+      "title": "Width & context ablation: loss saturation",
+      "layout": "full",
+      "figures": [
+        "figures/sec03_width_loss.png"
+      ],
+      "keywords": [
+        "Trunk x4: val loss −31% (0.197 → 0.136)",
+        "ctx x4, prop hint: loss = 1x",
+        "1x flat E8; trunk x4 still falling"
+      ],
+      "footnote": "Old 6k panel (1% floor); one validation set. 16x: memory-floor stop after part 2.",
+      "say": "All arms share the 6k panel (1% floor), corpus, seed and schedule; validation = 192 recipes x 64 cells. ctx x4, both x4 and 16x were read from server_2 snapshots.",
+      "sources": "tools/sec03_width_loss.py; data/sec03_width_loss.csv, data/sec03_width_saturation.csv, data/sec03_width_loss_headline.csv"
+    },
+    {
+      "kind": "content",
+      "section": "3",
+      "section_label": "3 Architecture ablation",
+      "title": "Width & context ablation: benchmark",
+      "layout": "full",
+      "figures": [
+        "figures/sec03_width_bench.png"
+      ],
+      "keywords": [
+        "trunk x4: loss −31%, no DEG gain",
+        "Prop hint: oracle = predicted, L1 worse",
+        "1x kept: ties ctx x4, 54M vs 102M"
+      ],
+      "footnote": "SLE, 42 donors, 2,980 genes; old 6k panel. ctx x4: better L1 (0.161 vs 0.197).",
+      "say": "Bars = mean of parts 14-16 with the cut p > 0.5; diamonds = Bernoulli read-out (4 width arms only). Prop hint = 9 L1 cell-type proportions; older panel, not the final model.",
+      "sources": "tools/sec03_width_bench.py; data/sec03_width_bench.csv, data/sec03_width_bench_summary.csv, data/sec03_width_bench_headline.csv"
+    },
+    {
+      "kind": "content",
+      "section": "3",
+      "section_label": "3 Architecture ablation",
+      "title": "Gene context: 2k / 3k / 6k / 15k",
+      "layout": "full",
+      "figures": [
+        "figures/sec03_panel.png"
+      ],
+      "keywords": [
+        "6k ≈ 3k: 3k inside 6k seed range",
+        "2k best L1; 15k worst r, L1",
+        "6k kept: first panel with CLEC9A (cDC1)"
+      ],
+      "footnote": "SLE, 42 donors, 1,006 benchmark genes in every panel; 6k base seed = best of 5.",
+      "say": "Runs f01_{2k,3k,6k,p15k}_1x, part 16, Bernoulli, EMA, DDIM 50. Different gene axis from the previous slide, so the two slides cannot be compared. Loss curves are divided by each panel's own part-16 value.",
+      "sources": "tools/sec03_panel.py; data/sec03_panel_bench.csv, data/sec03_panel_seeds.csv, data/sec03_panel_loss.csv, data/sec03_panel_headline.csv"
+    },
+    {
+      "kind": "content",
+      "section": "4",
+      "section_label": "4 SLE benchmark",
+      "title": "SLE benchmark: PBISC vs scSemiProfiler",
+      "layout": "full",
+      "figures": [
+        "figures/sec04_sle_benchmark.png"
+      ],
+      "keywords": [
+        "42 held-out donors, sc-pseudobulk input",
+        "Beats scSemiProfiler on all three metrics",
+        "Ties PBISC-VAE; below bulk-copy baseline"
+      ],
+      "footnote": "GSE135779, 42 held-out donors (30 SLE / 12 HC). PBISC: 5 seeds, mean ± SD; baselines: 1 draw.",
+      "sources": "tools/sec04_sle_benchmark.py; data/sec04_sle_benchmark.csv, data/sec04_sle_numbers.csv",
+      "footnote_full": "GSE135779, 42 held-out donors (30 SLE / 12 HC). PBISC f01 6k E16, 5 seeds (mean ± SD); baselines 1 draw. scSemiProfiler also gets real sc of 14 in-cohort donors (1,500 cells each). Dashed: input bulk logFC copied to every type (backup)."
+    },
+    {
+      "kind": "content",
+      "section": "4",
+      "section_label": "4 SLE benchmark",
+      "title": "SLE benchmark: per cell type",
+      "layout": "full",
+      "figures": [
+        "figures/sec04_sle_per_celltype.png"
+      ],
+      "keywords": [
+        "Beats scSemiProfiler in T, B, Mono, NK",
+        "Loses in rare DC and Other",
+        "DC: ~15 generated cells per donor"
+      ],
+      "footnote": "Same 42 donors, 1,006 genes. PBISC: 5 seeds, mean ± SD; baselines: 1 draw.",
+      "say": "PBISC's DC top list in the base seed has only 13 genes, because few DC cells are generated.",
+      "sources": "tools/sec04_sle_benchmark.py; data/sec04_sle_per_celltype.csv, data/sec04_sle_numbers.csv",
+      "footnote_full": "Same 42 donors, 1,006 genes. PBISC 5 seeds (mean ± SD), baselines 1 draw. Tick %: share of 251,179 real cells. Other top-50 is out of 12 truth genes. Bulk copy: r 0.68, top-50 113 (backup)."
+    },
+    {
+      "kind": "content",
+      "section": "5",
+      "section_label": "5 Real bulk vs pseudobulk",
+      "title": "Real bulk ≠ pseudobulk (model input)",
+      "layout": "figure_keywords",
+      "figures": [
+        "figures/sec05_condition_space.png"
+      ],
+      "keywords": [
+        "Real bulk: 2.46 D from own pseudobulk",
+        "95% of squared offset is one shift",
+        "Shift removed (diagnostic): own donor 7.5% → 75.7%"
+      ],
+      "footnote": "ImMicro, 107 healthy donors. Model input space (log1p CPM). D = median donor-to-donor distance.",
+      "say": "The two pseudobulk clusters on PC1 are sc pools DB1-3 vs DB4-7, so the pairing lines cross mainly because of sc batch. Shift removal subtracts the mean offset of the other 106 donors; a bulk-only deployment would not have them.",
+      "sources": "tools/sec05_condition_space.py; data/sec05_condition_space_numbers.csv, data/sec05_condition_space_coords.csv",
+      "footnote_full": "ImMicro, 107 healthy donors. Model inputs (log1p CPM, 5,508 genes), not learned ctx. D = median donor-to-donor distance; PCA-30 values are lower bounds (full space 3.19 D, 0.91 shared). Shift removal is a diagnostic that uses paired pseudobulks. Pseudobulk PC1 is sc pool batch (pool R² 0.96 on PC1)."
+    },
+    {
+      "kind": "content",
+      "section": "5",
+      "section_label": "5 Real bulk vs pseudobulk",
+      "title": "Consequence: cells off-manifold",
+      "layout": "two_figures",
+      "figures": [
+        "figures/sec05_cells_umap.png",
+        "figures/sec05_cells_metrics.png"
+      ],
+      "keywords": [
+        "Monocytes: raw 30.6% vs real 16.7%",
+        "Same noise: 19% T → monocyte",
+        "Precision/coverage: raw 0.36/0.08, pseudobulk 0.92/0.51"
+      ],
+      "footnote": "ImMicro, 107 donors; same starting noise for every input. refcal = cohort-mean bulk anchored to a healthy reference.",
+      "footnote_full": "ImMicro, 107 donors. Final model f01 6k 1× E16 (DDIM 50, EMA); each cell starts from the same noise under every input. Precision/coverage come from one scoring call. Composition L1: raw 0.455, pseudobulk 0.166, cohort-mean null 0.161. The UMAP is fitted on real cells and overstates overlap. refcal definition: plans/utility_studies_final6k_20260925/refcal/RESULTS.md ('Reference-anchored calibration ... anchoring a cohort's mean bulk to the AIFI healthy-PBMC reference'); AIDA reference: refcal_refs/RESULTS.md.",
+      "say": "TPM fixes cell shape (precision 0.87) but not monocytes (34.3%). refcal fixes monocytes (about 17.5%); coverage depends on the reference (AIFI 0.08, AIDA 0.37).",
+      "sources": "tools/sec05_cells_offmanifold.py; data/sec05_cells_metrics.csv, data/sec05_cells_umap_coords.csv"
+    },
+    {
+      "kind": "content",
+      "section": "5",
+      "section_label": "5 Real bulk vs pseudobulk",
+      "title": "Consequence: disease direction flips",
+      "layout": "figure_keywords",
+      "figures": [
+        "figures/sec05_behcet_direction.png"
+      ],
+      "keywords": [
+        "CD16 gap: raw −3.97 vs observed +4.94",
+        "C1Q flip vanishes at common depth",
+        "AIFI refcal restores sign, 5/5 seeds"
+      ],
+      "footnote": "Behçet. Bulk GSE198533 (9 BD / 10 HC); observed sc GSE198616 (4 / 4, different people).",
+      "say": "Limitations: AIDA refcal does not restore the sign (C1Q -0.25, CD16 +0.37). Best reference differs by cohort (AIDA for ImMicro, AIFI for Behcet). The sc pseudobulk ceiling uses the 8 sc donors, not the 19 bulk donors. Monocytes stay about 2x observed even under refcal.",
+      "sources": "tools/sec05_behcet_direction.py; data/sec05_behcet_direction.csv",
+      "footnote_full": "Behçet, final model. Bulk GSE198533 (BD 9, HC 10), 5,000 cells per sample; observed = sc GSE198616 (BD 4, HC 4, different people; descriptive p). At common depth, raw CD16 is −3.59 (seed 0) and raw C1Q is −0.08 ± 0.09 pp. AIDA refcal does not restore the sign."
+    },
+    {
+      "kind": "content",
+      "section": "5",
+      "section_label": "5 Real bulk vs pseudobulk",
+      "title": "Real bulk + refcal vs scSemiProfiler (ImMicro)",
+      "layout": "full",
+      "figures": [
+        "figures/sec04_immicro_external.png"
+      ],
+      "keywords": [
+        "Real bulk + refcal; no target sc",
+        "Lower L1, IFN error; rank not better",
+        "No-bulk replay: lower T error, rank ~0"
+      ],
+      "footnote": "ImMicro, 107 donors, 1 seed; CI = paired donor bootstrap.",
+      "say": "Section 4 author's slide ('External reference setting'), moved here after refcal is introduced. scSemiProfiler IFN variance ratio 0.036 (Mono) and 0.027 (T), vs PBISC 0.830 and 1.538.",
+      "sources": "tools/sec04_immicro.py; data/sec04_immicro_external.csv, data/sec04_immicro_numbers.csv",
+      "footnote_full": "ImMicro baseline, 107 donors, 1 generation seed, 5,500 genes; CI = paired donor bootstrap. Refcal is fit on the mean of all 107 bulks (transductive). PBISC pretraining includes part of AIFI and is not counted in generation time."
+    },
+    {
+      "kind": "content",
+      "section": "6",
+      "section_label": "6 Paired fine-tuning",
+      "title": "Fine-tuning with paired bulk + sc",
+      "layout": "diagram",
+      "figures": [
+        "figures/sec06_ft_method.png"
+      ],
+      "keywords": [
+        "LoRA 2–14 min vs pretraining 15.9 h"
+      ],
+      "footnote": "Base: final model f01 6k 1x E16. Times from 40 LoRA runs on this host.",
+      "sources": "tools/sec06_finetune_method.py; data/sec06_ft_method.csv",
+      "footnote_full": "Receipts: immicro/fold0/real_lora and ye2026_sepsis/fewshot_n5/draw0/lora RUN.json. Base: f01_6k_1x E16. Times: 40 local LoRA runs."
+    },
+    {
+      "kind": "content",
+      "section": "7",
+      "section_label": "7 LoRA ablations",
+      "title": "Full vs LoRA vs cond.-encoder-only",
+      "layout": "full",
+      "figures": [
+        "figures/sec07_ft_modes.png"
+      ],
+      "keywords": [
+        "LoRA ≈ full FT accuracy (n.s.)",
+        "42× fewer trainable params (1.3 M)",
+        "Cond.-encoder-only: no gain over zero-shot cal"
+      ],
+      "footnote": "Full vs LoRA: paired Wilcoxon on donor L1, p 0.25 / 0.71.",
+      "sources": "tools/sec07_ft_modes.py; data/sec07_ft_modes_accuracy.csv, data/sec07_ft_modes_compute.csv, data/sec07_ft_modes_wilcoxon.csv",
+      "footnote_full": "Base f01_6k_1x E16. Full vs LoRA: paired Wilcoxon, donor L1, p 0.25 / 0.71. Open = fold-0 pilot; compare same marker only. Per-step time: runs early-stop differently."
+    },
+    {
+      "kind": "content",
+      "section": "7",
+      "section_label": "7 LoRA ablations",
+      "title": "K diversification: no gain",
+      "layout": "figure_keywords",
+      "figures": [
+        "figures/sec07_k_aug.png"
+      ],
+      "keywords": [
+        "No gain; keep single K 1000",
+        "Multi-K slightly worse on ImMicro (p 0.04)",
+        "Inner-val loss Δ ≤ 1.2e-4"
+      ],
+      "footnote": "LoRA r16, 5-fold pooled. p = paired Wilcoxon on donor L1, uncorrected.",
+      "sources": "tools/sec07_k_aug.py; data/sec07_k_aug.csv, data/sec07_k_aug_wilcoxon.csv, data/sec07_k_aug_inner_val_delta.csv",
+      "footnote_full": "LoRA r16 on f01_6k_1x E16, 5-fold pooled. p: paired Wilcoxon, donor L1 vs plain LoRA, uncorrected. Recipes arm: K 1000 real bulk + ½-batch pretraining pseudobulk, no multi-K."
+    },
+    {
+      "kind": "content",
+      "section": "8",
+      "section_label": "8 Pairs needed",
+      "title": "Pairs needed: composition, coverage, training steps",
+      "layout": "full",
+      "figures": [
+        "figures/sec08_pairs_realism.png"
+      ],
+      "keywords": [
+        "Coverage, 5+7 pairs: ImMicro 0.84, sepsis 0.77"
+      ],
+      "footnote": "Donor 5-fold CV (ImMicro 107, sepsis 82 donors). N+7 = N training + 7 early-stop pairs.",
+      "footnote_full": "Final f01 6k·1× epoch-16 + LoRA; donor 5-fold CV, pooled scANVI scoring (ImMicro 107, sepsis 82 donors). N+7 = N training + 7 (sepsis 7–8) inner-val pairs for early stopping. cal = zero-shot + calibration, α fit on all 85–86 / 65–66 fold pairs; with α from the same 5 pairs (Sec. 9 call), coverage 0.54 vs LoRA 0.79 (ImMicro), 0.43 vs 0.69 (sepsis). Sec. 9 5-pair runs: fixed 3k steps, no inner-val. 20k cap reached by ImMicro folds 1/5 (10), 4/5 (20), 5/5 (40, all): val loss still falling, but last 2k steps <0.4% of total drop.",
+      "say": "Keywords say +7; 2 of 5 sepsis folds use 8 inner-val pairs.",
+      "sources": "tools/sec08_pairs_needed.py; data/sec08_pairs_realism.csv, data/sec08_selected_steps.csv, data/sec08_slide_numbers.csv"
+    },
+    {
+      "kind": "content",
+      "section": "8",
+      "section_label": "8 Pairs needed",
+      "title": "Pairs needed: donor and disease signal",
+      "layout": "figure_keywords",
+      "figures": [
+        "figures/sec08_pairs_signal.png"
+      ],
+      "keywords": [
+        "logFC r above broadcast from 10+7 pairs",
+        "DEG matches broadcast at 40+7 pairs",
+        "NK r 0.70 at 40+7 (cal 0.50)"
+      ],
+      "footnote": "Sepsis: 61 case vs 21 control. logFC r = mean over T / B / Mono / NK.",
+      "footnote_full": "Same runs and pooled scoring as previous slide; sepsis 61 case vs 21 control. N+7 = N training + 7 (sepsis 7–8) inner-val pairs. logFC r = mean over T/B/Mono/NK (mean CP10K); DEG = Welch top-50 overlap summed over T/B/Mono/NK/DC. ≤10 pairs: NK r (0.20/0.40) and DEG (5/8) below cal (0.50, 11); T r below cal until 40 pairs, Mono until all. DEG gain is Mono/DC; T overlap 5 (cal) → 0–1 from 20 pairs, NK 0 at every n (broadcast T 3, NK 2).",
+      "sources": "tools/sec08_pairs_needed.py; data/sec08_pairs_signal.csv, data/sec08_slide_numbers.csv"
+    },
+    {
+      "kind": "content",
+      "section": "9",
+      "section_label": "9 Paired validation",
+      "sec09_index": 0,
+      "title": "Five paired cohorts",
+      "layout": "full",
+      "figures": [
+        "figures/sec09_cohorts.png"
+      ],
+      "keywords": [
+        "403 paired samples, 275 donors",
+        "Only identity-pass pairs; donor-level splits",
+        "Ji bulk: whole blood, not PBMC"
+      ],
+      "footnote": "Usable pair = passed identity check and ≥ 200 cells: 403 of 430 pairs.",
+      "footnote_full": "Usable pair = passed the pairing identity check and >= 200 cells (fold >= 0 in data/<cohort>/folds.csv); dropped 15 identity-fail, 11 borderline and 1 low-cell pairs of 430. ImMicro: no identity flag, all used. Imperial, Giroux, Ji: several pairs per donor. Case vs control = samples in scores/<cohort>/pooled/summary.json. Giroux 'moderate' = GEO 'severe' (paper term). Labels: data/<cohort>/QC_REPORT.md; ImMicro: GEO GSE314922/GSE314416.",
+      "sources": "tools/sec09_paired_validation.py; data/sec09_cohorts.csv, data/sec09_quoted_numbers.csv, data/sec09_slide_text.json",
+      "numbers": [
+        {
+          "value": "403 paired samples (sum of usable rows)",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/<cohort>/fewshot_n5/DRAWS.json (n_usable_rows)"
+        },
+        {
+          "value": "275 donors (sum of usable donors)",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/<cohort>/fewshot_n5/DRAWS.json (n_usable_donors)"
+        },
+        {
+          "value": "dropped 15 identity-fail, 11 borderline, 1 < 200 cells, of 430",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/<cohort>/features.npz (identity_pass, identity_check); samples.csv (n_cells); /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/tools/prepare_dataset.py (MIN_CELLS)"
+        },
+        {
+          "value": "ImMicro 107 pairs / 107 donors",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/immicro/fewshot_n5/DRAWS.json"
+        },
+        {
+          "value": "Sepsis 82 pairs / 82 donors; sepsis 61 vs control 21",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/ye2026_sepsis/fewshot_n5/DRAWS.json; /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/ye2026_sepsis/pooled/summary.json"
+        },
+        {
+          "value": "Imperial 157 pairs / 57 donors; COVID+ 113 vs COVID− 34 (+10 recovery, neither)",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/imperial_eskd/fewshot_n5/DRAWS.json; /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/imperial_eskd/pooled/summary.json"
+        },
+        {
+          "value": "Giroux 24 pairs / 9 donors; moderate 11 vs mild 13",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/giroux2022/fewshot_n5/DRAWS.json; /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/giroux2022/pooled/summary.json"
+        },
+        {
+          "value": "Ji 33 pairs / 20 donors; ChILI 18 vs ICI control 15",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/ji2025_chili/fewshot_n5/DRAWS.json; /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/ji2025_chili/pooled/summary.json"
+        }
+      ]
+    },
+    {
+      "kind": "content",
+      "section": "9",
+      "section_label": "9 Paired validation",
+      "sec09_index": 1,
+      "title": "5 donors: LoRA vs scSemiProfiler — cells",
+      "layout": "full",
+      "figures": [
+        "figures/sec09_fewpair_cells.png"
+      ],
+      "keywords": [
+        "LoRA-5 > scSemi-5: coverage, precision 5/5; L1 4/5",
+        "Coverage: LoRA-5 0.69–0.96 vs scSemi-5 0.04–0.61",
+        "Composition gain vs zero-shot: ImMicro only"
+      ],
+      "footnote": "5 draws per cohort; test = all other usable pairs. scANVI labels; P/C = PCA-50, k = 10.",
+      "footnote_full": "Base f01 6k·1x·E16 (EMA, DDIM 50) + LoRA r=16, 3,000 steps. LoRA-5 / scSemi-5 = LoRA / scSemiProfiler trained on the same 5 donors (5–15 pairs); test = all other usable pairs. scSemi-5 scored in a separate call, same test set and reference (sha-checked); 25/25 draws (missing: none). Verdict = better in >= 80% of shared draws (L1: Giroux split 3/5). Depth-matched P/C: same ranking, coverage LoRA-5 0.83–0.97 vs scSemi-5 0.15–0.28. L1 vs zero-shot: per-donor Wilcoxon, ImMicro p 2e-10; others p >= 0.38. scANVI labels; P/C = PCA-50, k=10 balls. Scores: scores/<cohort>/fewshot_n5/.",
+      "say": "scSemi-5 is missing draw4 for ImMicro and Sepsis. When those arrive, rerun tools/sec09_paired_validation.py, then tools/sync_sec09_text.py, then generate_deck.py.",
+      "sources": "tools/sec09_paired_validation.py; data/sec09_fewpair_cells.csv, data/sec09_scsemi_status.csv, data/sec09_slide_text.json",
+      "numbers": [
+        {
+          "value": "LoRA-5 vs scSemi-5 coverage (shared-draw means; LoRA-better draws/n; verdict): ImMicro 0.792 vs 0.036 (5/5, lora); Sepsis 0.694 vs 0.080 (5/5, lora); Imperial 0.737 vs 0.059 (5/5, lora); Giroux 0.961 vs 0.607 (5/5, lora); Ji 0.707 vs 0.145 (5/5, lora) -> LoRA verdicts 5/5",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/<cohort>/fewshot_n5/draw<j>/summary.json and draw<j>_scsemi5/summary.json"
+        },
+        {
+          "value": "LoRA-5 vs scSemi-5 precision (shared-draw means; LoRA-better draws/n; verdict): ImMicro 0.952 vs 0.518 (5/5, lora); Sepsis 0.906 vs 0.728 (5/5, lora); Imperial 0.853 vs 0.330 (5/5, lora); Giroux 1.000 vs 0.992 (5/5, lora); Ji 0.978 vs 0.840 (5/5, lora) -> LoRA verdicts 5/5",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/<cohort>/fewshot_n5/draw<j>/summary.json and draw<j>_scsemi5/summary.json"
+        },
+        {
+          "value": "LoRA-5 vs scSemi-5 comp_l1 (shared-draw means; LoRA-better draws/n; verdict): ImMicro 0.176 vs 0.233 (5/5, lora); Sepsis 0.466 vs 0.612 (5/5, lora); Imperial 0.503 vs 0.602 (5/5, lora); Giroux 0.256 vs 0.270 (3/5, split); Ji 0.513 vs 0.592 (4/5, lora) -> LoRA verdicts 4/5",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/<cohort>/fewshot_n5/draw<j>/summary.json and draw<j>_scsemi5/summary.json"
+        },
+        {
+          "value": "LoRA-5 vs scSemi-5 coverage_dm (shared-draw means; LoRA-better draws/n; verdict): ImMicro 0.966 vs 0.284 (5/5, lora); Sepsis 0.834 vs 0.158 (5/5, lora); Imperial 0.863 vs 0.152 (5/5, lora); Giroux 0.974 vs 0.238 (5/5, lora); Ji 0.908 vs 0.150 (5/5, lora) -> LoRA verdicts 5/5",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/<cohort>/fewshot_n5/draw<j>/summary.json and draw<j>_scsemi5/summary.json"
+        },
+        {
+          "value": "LoRA-5 vs scSemi-5 precision_dm (shared-draw means; LoRA-better draws/n; verdict): ImMicro 0.975 vs 0.829 (5/5, lora); Sepsis 0.914 vs 0.504 (5/5, lora); Imperial 0.881 vs 0.429 (5/5, lora); Giroux 0.995 vs 0.426 (5/5, lora); Ji 0.974 vs 0.564 (5/5, lora) -> LoRA verdicts 5/5",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/<cohort>/fewshot_n5/draw<j>/summary.json and draw<j>_scsemi5/summary.json"
+        },
+        {
+          "value": "coverage cohort means over all draws: LoRA-5 ImMicro 0.792, Sepsis 0.694, Imperial 0.737, Giroux 0.961, Ji 0.707; scSemi-5 ImMicro 0.036, Sepsis 0.080, Imperial 0.059, Giroux 0.607, Ji 0.145",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/<cohort>/fewshot_n5/draw<j>/summary.json and draw<j>_scsemi5/summary.json"
+        },
+        {
+          "value": "LoRA-5 minus zero-shot-cal composition L1, per-donor Wilcoxon (draw-averaged): ImMicro -0.027, n=107, p=2.2e-10; Sepsis +0.018, n=82, p=0.79; Imperial +0.016, n=57, p=0.38; Giroux -0.065, n=8, p=0.38; Ji -0.012, n=20, p=0.7",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/<cohort>/fewshot_n5/draw<j>/per_donor.csv; /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/<cohort>/folds.csv"
+        },
+        {
+          "value": "scSemiProfiler draws used / LoRA draws: ImMicro 5/5, Sepsis 5/5, Imperial 5/5, Giroux 5/5, Ji 5/5",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/data/sec09_scsemi_status.csv"
+        },
+        {
+          "value": "training pairs per draw: ImMicro 5, Sepsis 5, Imperial 12–15, Giroux 12–14, Ji 7–9; test pairs: ImMicro 102, Sepsis 77, Imperial 142–145, Giroux 10–12, Ji 24–26",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/<cohort>/fewshot_n5/DRAWS.json (n_train_rows, n_test_rows)"
+        },
+        {
+          "value": "LoRA r = 16; 3000 fixed steps; base runs/f01_6k_1x_seed20260627/checkpoint_epoch_016.pt; sampling: test donors only, 1,500 cells each, seed 20260927, DDIM 50, EMA, Bernoulli read-out",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/runs/immicro/fewshot_n5/draw0/lora/RUN.json; /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/README.md"
+        },
+        {
+          "value": "P/C definition: precision/coverage/density (PCA-50 + k = 10 balls fitted on the scored donors' observed R1 cells); PCA components in every summary.json = 50",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/tools/score_ft.py"
+        }
+      ]
+    },
+    {
+      "kind": "content",
+      "section": "9",
+      "section_label": "9 Paired validation",
+      "sec09_index": 2,
+      "title": "5 donors: LoRA vs scSemiProfiler — disease signal",
+      "layout": "figure_keywords",
+      "figures": [
+        "figures/sec09_fewpair_disease.png"
+      ],
+      "keywords": [
+        "Top-50 DEG: LoRA-5 higher 3/4; Ji ~0",
+        "logFC r: scSemi-5 wins Giroux; Sepsis split",
+        "Giroux: real-bulk broadcast beats all arms"
+      ],
+      "footnote": "Same draws as previous slide. Top-50 DEG = Welch + BH overlap, summed over 5 cell types.",
+      "footnote_full": "Same draws and verdict rule as the previous slide; DEG tie = half the draws equal or mean gap < 1 gene. logFC r = mean over T/B/Mono/NK of case-vs-control log2FC r (generated vs observed); top-50 DEG = Welch+BH overlap summed over 5 types (max 250; Giroux 51–109, Ji 75–250 per draw). Giroux test = 2 vs 2 subjects per draw. Broadcast = the input real bulk's own logFC. Scores: scores/<cohort>/fewshot_n5/draw<j>{,_scsemi5}/.",
+      "sources": "tools/sec09_paired_validation.py; data/sec09_fewpair_disease.csv, data/sec09_slide_text.json",
+      "numbers": [
+        {
+          "value": "Top-50 DEG LoRA-5 vs scSemi-5 (shared-draw means; LoRA-better / tied draws; verdict): Sepsis 16.4 vs 5.2 (5/5, 0 tied, lora); Imperial 113.6 vs 46.8 (5/5, 0 tied, lora); Giroux 11.6 vs 1.2 (5/5, 0 tied, lora); Ji 0.2 vs 0.0 (1/5, 4 tied, tie)",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/<cohort>/fewshot_n5/draw<j>{,_scsemi5}/summary.json and contrast.csv"
+        },
+        {
+          "value": "logFC r LoRA-5 vs scSemi-5 (shared-draw means; LoRA-better draws; verdict): Sepsis 0.350 vs 0.402 (3/5, split); Imperial 0.749 vs 0.562 (5/5, lora); Giroux 0.331 vs 0.398 (0/5, scsemi); Ji 0.333 vs 0.153 (5/5, lora)",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/<cohort>/fewshot_n5/draw<j>{,_scsemi5}/summary.json and contrast.csv"
+        },
+        {
+          "value": "Giroux broadcast logFC r 0.539 vs best arm 0.398; broadcast DEG 25.4 vs best arm 11.6 (means over all draws)",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/giroux2022/fewshot_n5/draw<j>{,_scsemi5}/summary.json"
+        },
+        {
+          "value": "DEG maximum per draw: Sepsis 250, Imperial 250, Giroux 51–109, Ji 75–250",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/<cohort>/fewshot_n5/draw<j>/summary.json (deg_top50_max)"
+        },
+        {
+          "value": "Giroux test subjects per draw: 2 vs 2 (moderate vs mild); test pairs 10–12",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/giroux2022/fewshot_n5/DRAWS.json (train_donors); /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/giroux2022/folds.csv (group)"
+        }
+      ]
+    },
+    {
+      "kind": "content",
+      "section": "9",
+      "section_label": "9 Paired validation",
+      "sec09_index": 3,
+      "title": "All pairs, 5-fold",
+      "layout": "full",
+      "figures": [
+        "figures/sec09_fivefold.png"
+      ],
+      "keywords": [
+        "LoRA tops coverage 5/5, logFC r 4/4",
+        "Ji whole blood: baselines 0.11, LoRA 0.86",
+        "Giroux: broadcast r 0.64 > LoRA 0.50"
+      ],
+      "footnote": "Donor-level 5-fold CV; each pair scored by a model that never saw its donor.",
+      "footnote_full": "Donor-level 5-fold CV; each pair scored once, by the fold model that never trained on its donor (one pooled call per cohort). LoRA r=16 on all training donors' real bulk, early stop on inner-val donors; sc-only = full fine-tune on training donors' sc pseudobulk, calibrated input. Base f01 6k·1x·E16. Top-50 DEG: LoRA best 4/4. Sepsis DEG from pooled_deg. Scores: scores/<cohort>/pooled/summary.json.",
+      "sources": "tools/sec09_paired_validation.py; data/sec09_fivefold.csv, data/sec09_slide_text.json",
+      "numbers": [
+        {
+          "value": "LoRA best of 3 arms: coverage 5/5, logFC r 4/4, top-50 DEG 4/4",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/<cohort>/pooled/summary.json (Sepsis DEG: /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/ye2026_sepsis/pooled_deg/summary.json)"
+        },
+        {
+          "value": "Ji coverage: zero-shot cal 0.106, sc-only 0.113, LoRA 0.857",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/ji2025_chili/pooled/summary.json"
+        },
+        {
+          "value": "Giroux logFC r: broadcast 0.638 vs LoRA 0.504; DEG broadcast 39 vs LoRA 25",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/giroux2022/pooled/summary.json"
+        },
+        {
+          "value": "LoRA logFC r vs broadcast: Sepsis 0.527 vs 0.316, Imperial 0.819 vs 0.661, Giroux 0.504 vs 0.638, Ji 0.479 vs 0.240",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/<cohort>/pooled/summary.json (Sepsis DEG: /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/ye2026_sepsis/pooled_deg/summary.json)"
+        },
+        {
+          "value": "Pooled n (pairs / donors): ImMicro 107/107, Sepsis 82/82, Imperial 157/57, Giroux 24/9, Ji 33/20",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/<cohort>/pooled/summary.json (n_donors); /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/<cohort>/fewshot_n5/DRAWS.json"
+        },
+        {
+          "value": "early stopping: About 10 % of each fold's training donors are inner-validation, used only for early stopping",
+          "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/README.md"
+        }
+      ]
+    },
+    {
+      "kind": "content",
+      "section": "10",
+      "section_label": "10 Paired-subset cohorts",
+      "title": "Next: cohorts with a paired subset (planned)",
+      "layout": "figure_keywords",
+      "figures": [
+        "figures/sec10_paired_subset_cohorts.png"
+      ],
+      "keywords": [
+        "Train LoRA on pairs, generate bulk-only",
+        "Sepsis controls: 21 paired + 54 bulk-only",
+        "ChILI pre-ICI: 11 bulk, 4 unpaired sc"
+      ],
+      "footnote": "Counts from pairing manifests. Model: f01 6k 1x E16 + LoRA r16.",
+      "say": "Planned: no LoRA cells exist yet for any bulk-only sample. Ji bulk is PAXgene whole blood. Only 7 of the Arunachalam pairs are strict same-visit pairs. Asan is omitted because it has no donor map.",
+      "sources": "tools/sec10_paired_subset_cohorts.py; data/sec10_cohort_counts.csv, data/sec10_bulk_only_groups.csv, data/sec10_quoted_numbers.csv",
+      "footnote_full": "Counts from pairing manifests. QC = identity check (Imperial also >=200 cells); sex-flag = bulk sex differs from metadata or is ambiguous. Not shown: Asan (no donor map), Giroux (24 pairs, 17 bulk-only; broadcast beats LoRA). Model: f01 6k 1x E16 + LoRA r16."
+    },
+    {
+      "kind": "divider",
+      "title": "Backup"
+    },
+    {
+      "kind": "content",
+      "section": "4",
+      "section_label": "Backup · 4 SLE benchmark",
+      "backup": true,
+      "title": "Bulk broadcast baseline",
+      "layout": "full",
+      "figures": [
+        "figures/sec04_sle_broadcast.png"
+      ],
+      "keywords": [
+        "Bulk logFC copied: r 0.68 vs 0.37",
+        "Type-specific part: partial r 0.21",
+        "PBISC-VAE same; scSemiProfiler 0.17"
+      ],
+      "footnote": "SLE, 42 donors, 1,006 genes. Partial r: both logFCs residualised on bulk logFC.",
+      "sources": "tools/sec04_sle_benchmark.py; data/sec04_sle_broadcast.csv, data/sec04_sle_numbers.csv",
+      "footnote_full": "SLE held-out, 42 donors, 1,006 genes; PBISC 5 seeds. Partial r: truth and method logFC residualised on bulk logFC. Broadcast has no residual, so it is not scored. Split-half and BayesPrism checks: sle_broadcast/RESULTS.md."
+    },
+    {
+      "kind": "content",
+      "section": "5",
+      "section_label": "Backup · 5 Real bulk vs pseudobulk",
+      "backup": true,
+      "title": "Same shift in three paired cohorts",
+      "layout": "full",
+      "figures": [
+        "figures/sec05_shift_cohorts.png"
+      ],
+      "keywords": [
+        "Shared shift in all 3 cohorts"
+      ],
+      "footnote": "Model inputs only, no generation. PCA fitted per cohort on sc pseudobulks.",
+      "sources": "tools/sec05_shift_cohorts.py; data/sec05_shift_cohorts.csv",
+      "footnote_full": "Model inputs only, no generation. PCA is fitted on each cohort's sc pseudobulks (Arunachalam: 11 PCs, n = 12). D uses same-condition pairs only for Randolph and Arunachalam. Shift removal is a leave-one-out diagnostic."
+    },
+    {
+      "kind": "content",
+      "section": "6",
+      "section_label": "Backup · 6 Paired fine-tuning",
+      "backup": true,
+      "title": "Evaluation design",
+      "layout": "full",
+      "figures": [
+        "figures/sec06_eval_design.png"
+      ],
+      "keywords": [
+        "Donor-level splits; test donors never trained",
+        "5-fold: other folds · few: 5 donors",
+        "Giroux few-pair trains 12–14 of 24 pairs"
+      ],
+      "footnote": "Bars: Ye 2026 sepsis. QC-failed pairs excluded.",
+      "sources": "tools/sec06_finetune_method.py; data/sec06_eval_design.csv, data/sec06_eval_design_sepsis_roles.csv",
+      "footnote_full": "Counts: data/<cohort>/folds.csv and fewshot_n5/DRAWS.json; QC-failed pairs (fold −1) excluded. Bars: Ye 2026 sepsis."
+    }
+  ]
+}