Archive 4 checkpoints (1-4)

paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/scores/ji2025_chili/locoL_lora64sat_s51750_K1000/contrast.csv

@@ -0,0 +1,14 @@
+arm,celltype,statistic,n_genes,r_generated,r_broadcast,partial_r_generated_given_broadcast,r_generated_vs_broadcast,n_case_donors_obs,n_control_donors_obs,n_case_donors_gen,n_control_donors_gen,n_cells_obs,n_cells_gen,deg_scored,top50_overlap,top50_jaccard,obs_top_list_size,gen_top_list_size,n_sig_obs,n_sig_gen,top50_overlap_broadcast,top50_jaccard_broadcast,broadcast_top_list_size,n_sig_broadcast
+locoL_lora64sat_s51750_K1000,T cell,mean_cp10k,3021,0.23684360209538807,0.22039608693345034,0.20532012498494273,0.18093667119831655,18,15,18,15,59396,16361,,,,,,,,,,,
+locoL_lora64sat_s51750_K1000,T cell,pseudobulk_log1p_cpm,3021,0.22376706059148124,0.27145331834905573,0.2032830335340539,0.10777778823236063,18,15,18,15,59396,16361,,,,,,,,,,,
+locoL_lora64sat_s51750_K1000,B cell,mean_cp10k,3245,0.023477231182829895,0.13326750229861392,0.008832888464681974,0.11088309941608883,18,15,18,15,7162,9674,,,,,,,,,,,
+locoL_lora64sat_s51750_K1000,B cell,pseudobulk_log1p_cpm,3245,-0.0022843367336891534,0.01993963216355321,-0.0029560015093188074,0.0335718895762517,18,15,18,15,7162,9674,,,,,,,,,,,
+locoL_lora64sat_s51750_K1000,Monocyte,mean_cp10k,3154,0.26564185328116774,0.3983317004681723,0.17871048144388352,0.27073686293016064,18,15,18,15,34457,16843,,,,,,,,,,,
+locoL_lora64sat_s51750_K1000,Monocyte,pseudobulk_log1p_cpm,3154,0.15115779217774508,0.35676796217145185,0.10436867440129116,0.15364324383147027,18,15,18,15,34457,16843,,,,,,,,,,,
+locoL_lora64sat_s51750_K1000,NK cell,mean_cp10k,3051,0.13063094962289784,0.20750320008350984,0.11319332889451154,0.09850384428090714,18,15,18,15,18147,2833,,,,,,,,,,,
+locoL_lora64sat_s51750_K1000,NK cell,pseudobulk_log1p_cpm,3051,0.043439533935270024,0.10364196133517908,0.038084299000007114,0.05418637373690416,18,15,18,15,18147,2833,,,,,,,,,,,
+locoL_lora64sat_s51750_K1000,T cell,welch_top50_deg,5752,,,,,18,15,18,15,59396,16361,True,0,0.0,50,0,1897,0,0,0.0,15,15
+locoL_lora64sat_s51750_K1000,B cell,welch_top50_deg,5752,,,,,18,15,18,15,7162,9674,True,0,0.0,50,0,73,0,1,0.015625,15,15
+locoL_lora64sat_s51750_K1000,Monocyte,welch_top50_deg,5752,,,,,18,15,18,15,34457,16843,True,0,0.0,50,2,346,2,2,0.031746031746031744,15,15
+locoL_lora64sat_s51750_K1000,NK cell,welch_top50_deg,5752,,,,,18,15,18,15,18147,2833,True,0,0.0,50,3,396,3,1,0.015625,15,15
+locoL_lora64sat_s51750_K1000,DC,welch_top50_deg,5752,,,,,16,15,18,15,1755,725,True,0,0.0,50,0,138,0,0,0.0,15,15

paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/scores/ji2025_chili/locoL_lora64sat_s51750_K1000/per_donor.csv

@@ -0,0 +1,34 @@
+arm,donor,group,n_generated_cells,n_observed_cells,n_reference_cells,gen_T cell,obs_T cell,gen_B cell,obs_B cell,gen_Monocyte,obs_Monocyte,gen_NK cell,obs_NK cell,gen_DC,obs_DC,gen_Other,obs_Other,comp_l1,comp_l1_obs_loo_null,pb_r,pb_own_rank,pb_top1,precision,coverage,precision_depth_matched,coverage_depth_matched,mono_pct_gen,mono_pct_obs,median_umi_gen,median_umi_obs
+locoL_lora64sat_s51750_K1000,N064_TOL,TOL,1500,2498,150,0.33266666666666667,0.6517213771016813,0.18466666666666667,0.044035228182546036,0.38333333333333336,0.1677341873498799,0.037333333333333336,0.07606084867894315,0.016666666666666666,0.010008006405124099,0.04533333333333334,0.05044035228182546,0.7257784894582332,0.403363787668748,0.8351293204666944,30,False,0.9066666666666666,0.47333333333333333,0.88,0.4,38.333333333333336,16.77341873498799,1275.5,662.0
+locoL_lora64sat_s51750_K1000,N064_FU1,FU1,1500,4659,150,0.3953333333333333,0.6396222365314446,0.18333333333333332,0.016527151749302426,0.31333333333333335,0.13865636402661516,0.042666666666666665,0.19167203262502683,0.006666666666666667,0.00472204335694355,0.058666666666666666,0.008800171710667526,0.7865885383129428,0.4106007815223675,0.8413710492959007,16,False,0.92,0.18666666666666668,0.8933333333333333,0.35333333333333333,31.333333333333336,13.865636402661515,1307.0,925.0
+locoL_lora64sat_s51750_K1000,N032_AC,AC,1500,7350,150,0.33666666666666667,0.5891156462585034,0.27066666666666667,0.10979591836734694,0.2846666666666667,0.17551020408163265,0.050666666666666665,0.04707482993197279,0.015333333333333332,0.025714285714285714,0.042,0.0527891156462585,0.5472380952380953,0.40579421372527663,0.8452169962595278,11,False,0.9533333333333334,0.6733333333333333,0.94,0.44666666666666666,28.46666666666667,17.551020408163264,1526.5,437.0
+locoL_lora64sat_s51750_K1000,N038_BC,BC,1500,2801,150,0.31133333333333335,0.5608711174580507,0.212,0.0046411995715815784,0.33866666666666667,0.16565512317029632,0.059333333333333335,0.2170653338093538,0.016666666666666666,0.034273473759371655,0.062,0.017493752231345947,0.8497531833868857,0.3481378884463771,0.8411585962474709,20,False,0.98,0.47333333333333333,0.92,0.6466666666666666,33.86666666666667,16.565512317029633,1442.0,1317.0
+locoL_lora64sat_s51750_K1000,N038_AC,AC,1500,3976,150,0.2926666666666667,0.6830985915492958,0.21066666666666667,0.00829979879275654,0.37,0.1290241448692153,0.04666666666666667,0.12072434607645875,0.014,0.03194164989939638,0.066,0.026911468812877263,0.9648625083836352,0.4636716646911493,0.8502251276561938,10,False,0.96,0.7066666666666667,0.9333333333333333,0.68,37.0,12.90241448692153,1456.5,910.5
+locoL_lora64sat_s51750_K1000,N055_AC,AC,1500,4308,150,0.2753333333333333,0.340297121634169,0.19066666666666668,0.07799442896935933,0.38466666666666666,0.32474466109563604,0.06933333333333333,0.2051996285979573,0.012,0.0060352831940575676,0.068,0.0457288765088208,0.40166016713091923,0.3037845097207051,0.8473778014857818,8,False,0.94,0.6,0.9266666666666666,0.7133333333333334,38.46666666666667,32.47446610956361,1355.5,1139.5
+locoL_lora64sat_s51750_K1000,N136_TOL,TOL,1500,759,150,0.28933333333333333,0.48616600790513836,0.15533333333333332,0.10013175230566534,0.4066666666666667,0.19104084321475626,0.048666666666666664,0.0922266139657444,0.006666666666666667,0.0013175230566534915,0.09333333333333334,0.12911725955204217,0.5523530961791832,0.31649479655351276,0.8464890912900956,12,False,0.96,0.15333333333333332,0.8733333333333333,0.5333333333333333,40.666666666666664,19.104084321475625,1300.5,1837.0
+locoL_lora64sat_s51750_K1000,N136_FU1,FU1,1500,2927,150,0.266,0.4246668944311582,0.23133333333333334,0.14724974376494704,0.36666666666666664,0.19337205329689103,0.051333333333333335,0.09531943969935087,0.018666666666666668,0.0023915271609156134,0.066,0.13700034164673727,0.5473066848878259,0.41630882593211405,0.8586103514295933,4,False,0.9666666666666667,0.5666666666666667,0.9533333333333334,0.7333333333333333,36.666666666666664,19.3372053296891,1471.0,1501.0
+locoL_lora64sat_s51750_K1000,N138_TOL,TOL,1500,5129,150,0.2986666666666667,0.13803860401637746,0.19133333333333333,0.02807564827451745,0.382,0.5878338857477091,0.035333333333333335,0.17703255995320724,0.011333333333333334,0.006044063170208618,0.08133333333333333,0.06297523883798012,0.695066224735166,0.7925802602508564,0.8461941053328085,10,False,0.9533333333333334,0.38,0.9066666666666666,0.5333333333333333,38.2,58.78338857477091,1135.0,1117.0
+locoL_lora64sat_s51750_K1000,N138_FU1,FU1,1500,2107,150,0.22333333333333333,0.2728998576174656,0.14133333333333334,0.04841006169909824,0.484,0.4831514000949217,0.05466666666666667,0.1134314190792596,0.016666666666666666,0.005220692928334125,0.08,0.07688656858092074,0.21666255339345036,0.569899242948485,0.8574023254410559,3,False,0.98,0.37333333333333335,0.94,0.6666666666666666,48.4,48.31514000949217,1301.0,1446.0
+locoL_lora64sat_s51750_K1000,N128_BC,BC,1500,3902,150,0.33866666666666667,0.5158892875448488,0.19466666666666665,0.06817016914402871,0.3466666666666667,0.29420809841107126,0.04533333333333334,0.07842132239876987,0.023333333333333334,0.03331624807790876,0.051333333333333335,0.00999487442337263,0.44058704937638804,0.21132230546644887,0.8429053102782533,12,False,0.92,0.38,0.92,0.49333333333333335,34.66666666666667,29.420809841107125,1363.5,1088.0
+locoL_lora64sat_s51750_K1000,N128_AC,AC,1500,2713,150,0.296,0.5488389237007003,0.22066666666666668,0.06524143015112421,0.38666666666666666,0.2790269074824917,0.04133333333333333,0.07077036490969407,0.014,0.029487652045705862,0.04133333333333333,0.006634721710283818,0.595527214645534,0.23334452587077809,0.8506804881669753,7,False,0.9266666666666666,0.44,0.88,0.54,38.666666666666664,27.90269074824917,1311.0,1181.0
+locoL_lora64sat_s51750_K1000,N129_BC,BC,1500,3087,150,0.324,0.5963718820861678,0.226,0.11305474570780694,0.316,0.08357628765792031,0.05466666666666667,0.18140589569160998,0.025333333333333333,0.013281503077421444,0.054,0.012309685779073534,0.7982222222222223,0.42589167542025136,0.8511217438131871,5,False,0.9333333333333333,0.54,0.9466666666666667,0.7133333333333334,31.6,8.35762876579203,1468.5,1059.0
+locoL_lora64sat_s51750_K1000,N129_AC,AC,1500,2798,150,0.318,0.7076483202287348,0.236,0.04932094353109364,0.3453333333333333,0.07969978556111508,0.04133333333333333,0.13366690493209435,0.012666666666666666,0.0228734810578985,0.04666666666666667,0.006790564689063617,0.9843774124374552,0.4965696459014902,0.8462697758784372,14,False,0.9,0.6,0.8866666666666667,0.6266666666666667,34.53333333333333,7.969978556111508,1378.5,759.0
+locoL_lora64sat_s51750_K1000,N147_TOL,TOL,1500,3210,150,0.292,0.44205607476635517,0.17733333333333334,0.00809968847352025,0.35933333333333334,0.4193146417445483,0.09933333333333333,0.07102803738317758,0.014666666666666666,0.0065420560747663555,0.05733333333333333,0.0529595015576324,0.4200747663551402,0.36540380452880195,0.8754469109433042,1,True,0.98,0.6066666666666667,0.94,0.8733333333333333,35.93333333333334,41.93146417445483,1639.0,1461.5
+locoL_lora64sat_s51750_K1000,N150_BC,BC,1500,2433,150,0.37866666666666665,0.802301685162351,0.13733333333333334,0.02301685162351007,0.3466666666666667,0.1385121249486231,0.048,0.02178380600082203,0.016666666666666666,0.005754212905877518,0.07266666666666667,0.008631319358816275,0.8472700369913688,0.6657432679854348,0.8559868625015877,3,False,0.9466666666666667,0.5266666666666666,0.94,0.7933333333333333,34.66666666666667,13.851212494862311,1392.5,1629.0
+locoL_lora64sat_s51750_K1000,N158_AC,AC,1500,3749,150,0.30466666666666664,0.24619898639637236,0.18266666666666667,0.026673779674579887,0.38866666666666666,0.6295012003200854,0.064,0.04934649239797279,0.015333333333333332,0.024806615097359295,0.04466666666666667,0.0234729261136303,0.5006156308348894,0.7768443485656576,0.8534761325464183,9,False,0.9466666666666667,0.44,0.92,0.5666666666666667,38.86666666666667,62.95012003200854,1546.5,1047.0
+locoL_lora64sat_s51750_K1000,N159_BC,BC,1500,3779,150,0.5446666666666666,0.6996559936491136,0.15466666666666667,0.018523418893887273,0.158,0.08123842286319131,0.07866666666666666,0.173590897062715,0.017333333333333333,0.014024874305371792,0.04666666666666667,0.01296639322572109,0.4998271147569905,0.49076629176870806,0.8592012360404365,1,True,0.9733333333333334,0.7666666666666667,0.9333333333333333,0.8,15.8,8.123842286319132,1800.0,1045.0
+locoL_lora64sat_s51750_K1000,N001_TOL,TOL,1500,5658,150,0.3466666666666667,0.51926475786497,0.23,0.03375751148815836,0.306,0.21067515022976316,0.043333333333333335,0.2163308589607635,0.015333333333333332,0.003711558854718982,0.058666666666666666,0.016260162601626018,0.691191233651467,0.19394233264283423,0.8368925894669493,24,False,0.96,0.21333333333333335,0.9,0.3333333333333333,30.599999999999998,21.067515022976316,1384.5,787.5
+locoL_lora64sat_s51750_K1000,N001_FU1,FU1,1500,4547,150,0.2813333333333333,0.29601935341983726,0.21333333333333335,0.06575764240158347,0.33066666666666666,0.48823400043985044,0.09066666666666667,0.13591378931163403,0.009333333333333334,0.0010996261271167802,0.07466666666666667,0.012975588299978007,0.4350009530093102,0.48025199143000386,0.838254506079393,28,False,0.9666666666666667,0.1,0.9333333333333333,0.2866666666666667,33.06666666666666,48.823400043985046,1459.0,1165.0
+locoL_lora64sat_s51750_K1000,N002_TOL,TOL,1500,6318,150,0.3353333333333333,0.41579613801836024,0.21066666666666667,0.05777144666033555,0.33866666666666667,0.3779677113010446,0.037333333333333336,0.1179170623615068,0.03,0.020101297879075657,0.048,0.010446343779677113,0.40069515669515676,0.26738927826214887,0.8353448429004401,24,False,0.94,0.15333333333333332,0.9466666666666667,0.29333333333333333,33.86666666666667,37.79677113010446,1334.0,933.5
+locoL_lora64sat_s51750_K1000,N002_FU1,FU1,1500,4975,150,0.38,0.31618090452261305,0.18066666666666667,0.07276381909547738,0.3253333333333333,0.45587939698492463,0.05466666666666667,0.12261306532663317,0.017333333333333333,0.02271356783919598,0.042,0.00984924623115578,0.40774539363484097,0.4584219344204332,0.8495005855077603,7,False,0.9066666666666666,0.26,0.8533333333333334,0.3466666666666667,32.53333333333333,45.58793969849246,1387.0,979.0
+locoL_lora64sat_s51750_K1000,N005_BC,BC,1500,6359,150,0.3526666666666667,0.7200817738638151,0.15133333333333332,0.05016512030193427,0.37133333333333335,0.11448340934109137,0.054,0.09891492373014625,0.012666666666666666,0.007705614090265765,0.058,0.008649158672747287,0.8246600618545894,0.5179762638363822,0.853540753936913,7,False,0.9466666666666667,0.5666666666666667,0.9133333333333333,0.5266666666666666,37.13333333333333,11.448340934109137,1324.5,782.0
+locoL_lora64sat_s51750_K1000,N005_AC,AC,1500,3433,150,0.43933333333333335,0.8100786484124672,0.23533333333333334,0.015147101660355374,0.21266666666666667,0.07952228371686572,0.046,0.08243518788231867,0.013333333333333334,0.010195164579085349,0.05333333333333334,0.0026216137489076608,0.8143610059229052,0.6979700129336864,0.8530478398086745,12,False,0.9266666666666666,0.5666666666666667,0.88,0.84,21.266666666666666,7.952228371686572,1446.5,1126.0
+locoL_lora64sat_s51750_K1000,N012_TOL,TOL,1500,6709,150,0.29533333333333334,0.34759278581010583,0.15666666666666668,0.07884930690117753,0.4206666666666667,0.4672827545088687,0.042,0.08570576837084513,0.020666666666666667,0.0005962140408406618,0.06466666666666666,0.01997317036816217,0.28516261737963927,0.4808462270666366,0.8297869428735832,28,False,0.94,0.12,0.92,0.32666666666666666,42.06666666666667,46.72827545088687,1293.0,1193.0
+locoL_lora64sat_s51750_K1000,N012_FU1,FU1,1500,4369,150,0.29133333333333333,0.35889219501029984,0.15933333333333333,0.19249256122682537,0.4126666666666667,0.3126573586633097,0.048666666666666664,0.10849164568551156,0.012666666666666666,0.000686655985351339,0.07533333333333334,0.02677958342870222,0.3210861371786069,0.3988819440268142,0.8368921088274994,21,False,0.9466666666666667,0.22666666666666666,0.9266666666666666,0.6133333333333333,41.266666666666666,31.265735866330967,1250.0,1299.0
+locoL_lora64sat_s51750_K1000,N018_BC,BC,1500,3045,150,0.35533333333333333,0.32939244663382594,0.2846666666666667,0.042692939244663386,0.23333333333333334,0.2889983579638752,0.06866666666666667,0.2653530377668309,0.014666666666666666,0.05812807881773399,0.043333333333333335,0.015435139573070607,0.5916256157635468,0.3644697123957184,0.8443949409621162,20,False,0.98,0.52,0.96,0.58,23.333333333333332,28.899835796387517,1732.5,995.0
+locoL_lora64sat_s51750_K1000,N018_AC,AC,1500,3520,150,0.41,0.5900568181818182,0.198,0.03664772727272727,0.26,0.16164772727272728,0.058666666666666666,0.14914772727272727,0.012,0.04261363636363636,0.06133333333333333,0.019886363636363636,0.6023030303030305,0.2802228100673128,0.8610074924398138,2,False,0.9733333333333334,0.7933333333333333,0.9533333333333334,0.82,26.0,16.164772727272727,1459.5,858.0
+locoL_lora64sat_s51750_K1000,N022_TOL,TOL,1500,3651,150,0.2833333333333333,0.4864420706655711,0.20133333333333334,0.03697617091207888,0.3626666666666667,0.045740892906053134,0.062,0.4182415776499589,0.01,0.0,0.08066666666666666,0.01259928786633799,1.1187006299643936,0.562781565138013,0.8238377196034121,33,False,0.9466666666666667,0.07333333333333333,0.9133333333333333,0.38666666666666666,36.266666666666666,4.574089290605313,1375.5,1410.0
+locoL_lora64sat_s51750_K1000,N022_FU1,FU1,1500,2463,150,0.39466666666666667,0.48477466504263095,0.16933333333333334,0.048721071863580996,0.2906666666666667,0.20259845716605765,0.058666666666666666,0.25334957369062117,0.010666666666666666,0.0,0.076,0.010556232237109216,0.5695818107998376,0.22966274597345715,0.8330068365354381,28,False,0.9533333333333334,0.11333333333333333,0.94,0.5333333333333333,29.06666666666667,20.259845716605763,1364.5,1459.0
+locoL_lora64sat_s51750_K1000,N035_TOL,TOL,1500,1598,150,0.348,0.2853566958698373,0.20066666666666666,0.016896120150187734,0.26466666666666666,0.31101376720901125,0.09933333333333333,0.3404255319148936,0.011333333333333334,0.01376720901126408,0.076,0.03254067584480601,0.5797463496036712,0.4807456196784965,0.8434707944760643,23,False,0.9733333333333334,0.14,0.9333333333333333,0.7533333333333333,26.466666666666665,31.101376720901126,1653.0,1855.5
+locoL_lora64sat_s51750_K1000,N043_TOL,TOL,1500,2074,150,0.308,0.3548698167791707,0.164,0.05737704918032787,0.366,0.04580520732883317,0.068,0.4811957569913211,0.007333333333333333,0.0028929604628736743,0.08666666666666667,0.05785920925747348,0.9201311475409835,0.7311197263337292,0.8312283471224456,32,False,0.92,0.13333333333333333,0.8666666666666667,0.6533333333333333,36.6,4.580520732883317,1334.5,1315.5
+locoL_lora64sat_s51750_K1000,N043_FU1,FU1,1500,3437,150,0.268,0.1638056444573756,0.19333333333333333,0.06779167878964212,0.372,0.5728833284841431,0.088,0.14896712249054408,0.012,0.0034914169333721268,0.06666666666666667,0.043060808844922895,0.5237009019493745,0.6824679208462441,0.8510946622078205,19,False,0.9266666666666666,0.17333333333333334,0.9333333333333333,0.47333333333333333,37.2,57.28833284841431,1474.0,1422.0

paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/scores/ji2025_chili/locoL_lora64sat_s51750_K1000/summary.json

@@ -0,0 +1,569 @@
+{
+ "schema": "paired_ft_scores.v1",
+ "dataset": "ji2025_chili",
+ "written_utc": "2026-10-02T21:56:37.581470+00:00",
+ "command": [
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/tools/score_ft.py",
+  "--dataset",
+  "ji2025_chili",
+  "--threads",
+  "8",
+  "--donors",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/data/ji2025_chili/usable_samples.txt",
+  "--arm",
+  "locoL_lora64sat_s51750_K1000=/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/generated/ji2025_chili/hoL_ji2025_chili/K1000_lora64sat_s51750.npz",
+  "--case-groups",
+  "TOL,FU1",
+  "--control-groups",
+  "BC,AC",
+  "--out",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/scores/ji2025_chili/locoL_lora64sat_s51750_K1000"
+ ],
+ "arms": [
+  "locoL_lora64sat_s51750_K1000"
+ ],
+ "donors": [
+  "N064_TOL",
+  "N064_FU1",
+  "N032_AC",
+  "N038_BC",
+  "N038_AC",
+  "N055_AC",
+  "N136_TOL",
+  "N136_FU1",
+  "N138_TOL",
+  "N138_FU1",
+  "N128_BC",
+  "N128_AC",
+  "N129_BC",
+  "N129_AC",
+  "N147_TOL",
+  "N150_BC",
+  "N158_AC",
+  "N159_BC",
+  "N001_TOL",
+  "N001_FU1",
+  "N002_TOL",
+  "N002_FU1",
+  "N005_BC",
+  "N005_AC",
+  "N012_TOL",
+  "N012_FU1",
+  "N018_BC",
+  "N018_AC",
+  "N022_TOL",
+  "N022_FU1",
+  "N035_TOL",
+  "N043_TOL",
+  "N043_FU1"
+ ],
+ "n_donors": 33,
+ "reference": {
+  "cache": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ji2025_chili",
+  "reference_sha256": "19c30f56db699eb4f5a642927c1cd6e173212b877c29ff20d84e60a5ecccef6d",
+  "features_sha256": "049795e3a2470a223347015c0c965f56ca90c21143aad6f72d12e39f3e95e58b",
+  "n_r1_cells": 4950,
+  "pca": {
+   "n_components": 50,
+   "explained_variance_ratio_sum": 0.2840818762779236,
+   "n_fit_cells": 4950
+  },
+  "r2_control": {
+   "precision": 0.9961616161616161,
+   "coverage": 1.0,
+   "density": 2.2093535353535354,
+   "n_query": 4950,
+   "depth_matched": {
+    "precision": 0.9967676767676767,
+    "coverage": 1.0,
+    "density": 2.1805656565656566
+   }
+  },
+  "observed_l1_loo_null_mean": 0.45223248248542647
+ },
+ "contrast_definition": {
+  "enabled": true,
+  "rule": "explicit --case-groups / --control-groups",
+  "group_source": "features groups",
+  "case_groups": [
+   "TOL",
+   "FU1"
+  ],
+  "control_groups": [
+   "BC",
+   "AC"
+  ],
+  "n_case_donors": 18,
+  "n_control_donors": 15,
+  "n_donors_in_neither": 0,
+  "case": [
+   true,
+   true,
+   false,
+   false,
+   false,
+   false,
+   true,
+   true,
+   true,
+   true,
+   false,
+   false,
+   false,
+   false,
+   true,
+   false,
+   false,
+   false,
+   true,
+   true,
+   true,
+   true,
+   false,
+   false,
+   true,
+   true,
+   false,
+   false,
+   true,
+   true,
+   true,
+   true,
+   true
+  ],
+  "ctrl": [
+   false,
+   false,
+   true,
+   true,
+   true,
+   true,
+   false,
+   false,
+   false,
+   false,
+   true,
+   true,
+   true,
+   true,
+   false,
+   true,
+   true,
+   true,
+   false,
+   false,
+   false,
+   false,
+   true,
+   true,
+   false,
+   false,
+   true,
+   true,
+   false,
+   false,
+   false,
+   false,
+   false
+  ],
+  "groups": [
+   "TOL",
+   "FU1",
+   "AC",
+   "BC",
+   "AC",
+   "AC",
+   "TOL",
+   "FU1",
+   "TOL",
+   "FU1",
+   "BC",
+   "AC",
+   "BC",
+   "AC",
+   "TOL",
+   "BC",
+   "AC",
+   "BC",
+   "TOL",
+   "FU1",
+   "TOL",
+   "FU1",
+   "BC",
+   "AC",
+   "TOL",
+   "FU1",
+   "BC",
+   "AC",
+   "TOL",
+   "FU1",
+   "TOL",
+   "TOL",
+   "FU1"
+  ]
+ },
+ "deg_definition": {
+  "enabled": true,
+  "score": "top-50 DEG overlap per cell type (benchmark total_top50_overlap, Welch backend)",
+  "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/bulk2sc-paper/scripts/evaluate_generated_vs_truth.py",
+  "functions": [
+   "welch_bh_deg",
+   "top_deg_genes",
+   "jaccard"
+  ],
+  "reproduces": "utility_studies_final6k_20260925/sle top-50 DEG 81 (1,006 genes) / 84 (2,279 genes) = sum over the six gated types of top50_overlap_count",
+  "top_n": 50,
+  "fdr": 0.05,
+  "n_genes": 5752,
+  "types": [
+   "T cell",
+   "B cell",
+   "Monocyte",
+   "NK cell",
+   "DC"
+  ],
+  "aggregates_source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ji2025_chili/deg_reference.npz",
+  "aggregates_sha256": "4f5eb8c8389ef1308ce3ec44be8be024cedb97de37351d109d9070c818df5a9e",
+  "aggregates_note": null,
+  "types_not_scored_observed": [],
+  "observed": {
+   "T cell": {
+    "n_significant": 1897,
+    "n_tested": 5696,
+    "n_case_donors": 18,
+    "n_control_donors": 15,
+    "n_cells": 59396,
+    "top_list_size": 50
+   },
+   "B cell": {
+    "n_significant": 73,
+    "n_tested": 5418,
+    "n_case_donors": 18,
+    "n_control_donors": 15,
+    "n_cells": 7162,
+    "top_list_size": 50
+   },
+   "Monocyte": {
+    "n_significant": 346,
+    "n_tested": 5699,
+    "n_case_donors": 18,
+    "n_control_donors": 15,
+    "n_cells": 34457,
+    "top_list_size": 50
+   },
+   "NK cell": {
+    "n_significant": 396,
+    "n_tested": 5516,
+    "n_case_donors": 18,
+    "n_control_donors": 15,
+    "n_cells": 18147,
+    "top_list_size": 50
+   },
+   "DC": {
+    "n_significant": 138,
+    "n_tested": 5168,
+    "n_case_donors": 16,
+    "n_control_donors": 15,
+    "n_cells": 1755,
+    "top_list_size": 50
+   }
+  },
+  "broadcast": "welch_bh_deg + top_deg_genes on the arm's condition bulk_counts (scoring genes, case|control donors), one list compared with every type's observed list (sle_broadcast.py broadcast_input_bulk); not part of the published benchmark"
+ },
+ "headline": {
+  "locoL_lora64sat_s51750_K1000": {
+   "comp_l1": 0.6199837283023538,
+   "comp_l1_obs_loo_null": 0.45223248248542647,
+   "precision": 0.9470707070707071,
+   "coverage": 0.39515151515151514,
+   "precision_dm": 0.9183838383838384,
+   "coverage_dm": 0.5721212121212121,
+   "pb_r_median": 0.8462697758784372,
+   "pb_top1": 0.06060606060606061,
+   "mono_pct_gen": 34.02626262626263,
+   "mono_pct_obs": 26.35512163470612,
+   "median_umi_gen": 1412.0,
+   "median_umi_obs": 1056.0,
+   "contrast_mean_cp10k_r_gen": 0.16414840904557088,
+   "contrast_pseudobulk_log1p_cpm_r_gen": 0.10402001249270179,
+   "contrast_mean_cp10k_r_broadcast": 0.2398746224459366,
+   "contrast_pseudobulk_log1p_cpm_r_broadcast": 0.18795071850481,
+   "deg_top50_T": 0,
+   "deg_top50_B": 0,
+   "deg_top50_Mono": 0,
+   "deg_top50_NK": 0,
+   "deg_top50_DC": 0,
+   "deg_top50_sum": 0,
+   "deg_top50_max": 250,
+   "deg_top50_sum_4type": 0,
+   "deg_top50_sum_broadcast": 4,
+   "seconds": 14.2
+  }
+ },
+ "per_arm": {
+  "locoL_lora64sat_s51750_K1000": {
+   "arm": "locoL_lora64sat_s51750_K1000",
+   "npz": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/generated/ji2025_chili/hoL_ji2025_chili/K1000_lora64sat_s51750.npz",
+   "npz_sha256": "77e48a26c5468564f7e281c511bd2f4a11b1f311297af22fb095e6513d5e132f",
+   "receipt_verified": true,
+   "generation": {
+    "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ji2025_chili/K1000/lora64sat/checkpoint_ft_s51750.pt",
+    "checkpoint_sha256": "c6d0e7a0ec782d365f3eb32a958f76dc905e94bf720b2b86f0ab8f663a4bbe1f",
+    "input_condition_path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/data/ji2025_chili/test_raw_K1000.npz",
+    "input_condition_sha256": "78d8ac589dee9315807163c174fafde76bd890710e96c512773dfaf9151e5b86",
+    "seed": 20260927,
+    "steps": 50,
+    "support_readout": "bernoulli",
+    "weights": "ema"
+   },
+   "condition_for_broadcast": {
+    "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/data/ji2025_chili/test_raw_K1000.npz",
+    "sha256": "78d8ac589dee9315807163c174fafde76bd890710e96c512773dfaf9151e5b86"
+   },
+   "poisson": {
+    "seed": 20260916,
+    "rule": "whole NPZ, rows in NPZ order",
+    "nnz_whole_npz": 10977230,
+    "median_library_whole_npz_panel": 1412.0,
+    "n_rows_whole_npz": 49500
+   },
+   "annotation": {
+    "model_pt_sha256": "bf3655899e5abee435340fc008656c4a17342da0f30b70cfd43de8e9223e1930",
+    "zero_padded_model_genes": 1,
+    "label_counts": {
+     "Monocyte": 16843,
+     "T cell": 16361,
+     "B cell": 9674,
+     "Other": 3064,
+     "NK cell": 2833,
+     "DC": 725
+    },
+    "mean_l2_max_prob": 0.9137853215171231,
+    "predict_seconds": 6.0
+   },
+   "n_donors": 33,
+   "n_generated_cells": 49500,
+   "composition": {
+    "l1_mean": 0.6199837283023538,
+    "l1_median": 0.5797463496036712,
+    "l1_obs_loo_null_mean": 0.45223248248542647,
+    "l1_generated_loo_mean_vs_observed_mean": 0.6429044781219166,
+    "composition_top1": 0.030303030303030304,
+    "mean_fraction_generated": {
+     "T cell": 0.330525,
+     "B cell": 0.195434,
+     "Monocyte": 0.340263,
+     "NK cell": 0.057232,
+     "DC": 0.014646,
+     "Other": 0.061899
+    },
+    "mean_fraction_observed": {
+     "T cell": 0.479517,
+     "B cell": 0.055487,
+     "Monocyte": 0.263551,
+     "NK cell": 0.155783,
+     "DC": 0.014105,
+     "Other": 0.031557
+    }
+   },
+   "pseudobulk": {
+    "n_genes": 5752,
+    "r_median": 0.8462697758784372,
+    "r_mean": 0.8462292178127893,
+    "top1": 0.06060606060606061,
+    "n_top1": 2,
+    "own_rank_median": 12.0,
+    "chance_top1": 0.030303030303030304
+   },
+   "precision_coverage": {
+    "precision": 0.9470707070707071,
+    "coverage": 0.39515151515151514,
+    "density": 1.0387474747474748,
+    "n_query": 4950,
+    "n_reference": 4950
+   },
+   "precision_coverage_depth_matched": {
+    "precision": 0.9183838383838384,
+    "coverage": 0.5721212121212121,
+    "density": 0.9603030303030303,
+    "detected_median_before": {
+     "R1": 358.5,
+     "Q": 224.0
+    },
+    "detected_median_after": {
+     "R1": 223.0,
+     "Q": 223.0
+    }
+   },
+   "monocyte_pct": {
+    "generated_mean_over_donors": 34.02626262626263,
+    "observed_mean_over_donors": 26.35512163470612,
+    "generated_pooled_cells": 34.02626262626263
+   },
+   "median_umi_per_cell_scoring_genes": {
+    "generated_pooled": 1412.0,
+    "observed_pooled": 1056.0,
+    "generated_mean_of_donor_medians": 1416.4241943359375,
+    "observed_mean_of_donor_medians": 1155.8030303030303
+   },
+   "contrast": {
+    "mean_cp10k": {
+     "per_type": {
+      "T cell": {
+       "r_generated": 0.23684360209538807,
+       "r_broadcast": 0.22039608693345034,
+       "partial_r_generated_given_broadcast": 0.20532012498494273,
+       "r_generated_vs_broadcast": 0.18093667119831655,
+       "n_genes": 3021
+      },
+      "B cell": {
+       "r_generated": 0.023477231182829895,
+       "r_broadcast": 0.13326750229861392,
+       "partial_r_generated_given_broadcast": 0.008832888464681974,
+       "r_generated_vs_broadcast": 0.11088309941608883,
+       "n_genes": 3245
+      },
+      "Monocyte": {
+       "r_generated": 0.26564185328116774,
+       "r_broadcast": 0.3983317004681723,
+       "partial_r_generated_given_broadcast": 0.17871048144388352,
+       "r_generated_vs_broadcast": 0.27073686293016064,
+       "n_genes": 3154
+      },
+      "NK cell": {
+       "r_generated": 0.13063094962289784,
+       "r_broadcast": 0.20750320008350984,
+       "partial_r_generated_given_broadcast": 0.11319332889451154,
+       "r_generated_vs_broadcast": 0.09850384428090714,
+       "n_genes": 3051
+      }
+     },
+     "mean_r_generated": 0.16414840904557088,
+     "mean_r_broadcast": 0.2398746224459366,
+     "mean_partial_r": 0.12651420594700494
+    },
+    "pseudobulk_log1p_cpm": {
+     "per_type": {
+      "T cell": {
+       "r_generated": 0.22376706059148124,
+       "r_broadcast": 0.27145331834905573,
+       "partial_r_generated_given_broadcast": 0.2032830335340539,
+       "r_generated_vs_broadcast": 0.10777778823236063,
+       "n_genes": 3021
+      },
+      "B cell": {
+       "r_generated": -0.0022843367336891534,
+       "r_broadcast": 0.01993963216355321,
+       "partial_r_generated_given_broadcast": -0.0029560015093188074,
+       "r_generated_vs_broadcast": 0.0335718895762517,
+       "n_genes": 3245
+      },
+      "Monocyte": {
+       "r_generated": 0.15115779217774508,
+       "r_broadcast": 0.35676796217145185,
+       "partial_r_generated_given_broadcast": 0.10436867440129116,
+       "r_generated_vs_broadcast": 0.15364324383147027,
+       "n_genes": 3154
+      },
+      "NK cell": {
+       "r_generated": 0.043439533935270024,
+       "r_broadcast": 0.10364196133517908,
+       "partial_r_generated_given_broadcast": 0.038084299000007114,
+       "r_generated_vs_broadcast": 0.05418637373690416,
+       "n_genes": 3051
+      }
+     },
+     "mean_r_generated": 0.10402001249270179,
+     "mean_r_broadcast": 0.18795071850481,
+     "mean_partial_r": 0.08569500135650834
+    }
+   },
+   "deg_top50": {
+    "types_scored": [
+     "T cell",
+     "B cell",
+     "Monocyte",
+     "NK cell",
+     "DC"
+    ],
+    "types_not_scored": {},
+    "per_type": {
+     "T cell": {
+      "deg_scored": true,
+      "top50_overlap": 0,
+      "top50_jaccard": 0.0,
+      "obs_top_list_size": 50,
+      "gen_top_list_size": 0,
+      "top50_overlap_broadcast": 0,
+      "top50_jaccard_broadcast": 0.0,
+      "n_cells_obs": 59396,
+      "n_cells_gen": 16361
+     },
+     "B cell": {
+      "deg_scored": true,
+      "top50_overlap": 0,
+      "top50_jaccard": 0.0,
+      "obs_top_list_size": 50,
+      "gen_top_list_size": 0,
+      "top50_overlap_broadcast": 1,
+      "top50_jaccard_broadcast": 0.015625,
+      "n_cells_obs": 7162,
+      "n_cells_gen": 9674
+     },
+     "Monocyte": {
+      "deg_scored": true,
+      "top50_overlap": 0,
+      "top50_jaccard": 0.0,
+      "obs_top_list_size": 50,
+      "gen_top_list_size": 2,
+      "top50_overlap_broadcast": 2,
+      "top50_jaccard_broadcast": 0.031746031746031744,
+      "n_cells_obs": 34457,
+      "n_cells_gen": 16843
+     },
+     "NK cell": {
+      "deg_scored": true,
+      "top50_overlap": 0,
+      "top50_jaccard": 0.0,
+      "obs_top_list_size": 50,
+      "gen_top_list_size": 3,
+      "top50_overlap_broadcast": 1,
+      "top50_jaccard_broadcast": 0.015625,
+      "n_cells_obs": 18147,
+      "n_cells_gen": 2833
+     },
+     "DC": {
+      "deg_scored": true,
+      "top50_overlap": 0,
+      "top50_jaccard": 0.0,
+      "obs_top_list_size": 50,
+      "gen_top_list_size": 0,
+      "top50_overlap_broadcast": 0,
+      "top50_jaccard_broadcast": 0.0,
+      "n_cells_obs": 1755,
+      "n_cells_gen": 725
+     }
+    },
+    "sum": 0,
+    "sum_4type": 0,
+    "max": 250,
+    "mean": 0.0,
+    "sum_broadcast": 4,
+    "sum_broadcast_4type": 4,
+    "broadcast_n_significant": 15,
+    "seconds": 0.14
+   },
+   "seconds": 14.2
+  }
+ },
+ "code_sha256": {
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/tools/score_ft.py": "ec5f40b4cea6d5062d0138f3b934abb6465e34cbfeac869c8a0906924c033aec",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/tools/ft_common.py": "4a6366b0f22208cd58714fc0261b8840c643b9d4569704d560cfd93534cbc747",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/utility_studies_final6k_20260925/umap/cell_immicro_5arm/tools/stage_b_embed.py": "d056cfdf38fb0e5f478756f370174e993b75d0c973c48e02e21749175a4ca2da",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/bulk2sc-paper/scripts/evaluate_generated_vs_truth.py": "48b12d794782b840838853e93e0fc55fab00532c4b3698f4588e926b985d2881",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/utility_studies_final6k_20260925/immicro/tools/immicro_common.py": "c5c30e917148cc5e664e91ea30c11bbdff9b6d1f53a3e0a13e16b1917eff37e9",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/aifi_scanvi_rebuild_20260916/predict_scanvi.py": "482ee5dcf4d72b54bc36cbb6dcff62e5030bc7afc46cb5fb01eca73a82ee45a5",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/aifi_scanvi_rebuild_20260916/poisson_readout.py": "8bfa080942fbcdc89ae23b276524a1a83266fb8cbc56b9af279e50f82e6cde38"
+ },
+ "seconds": 16.8
+}

paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/scores/ji2025_chili/xcalL_K1000/README.md

@@ -0,0 +1,73 @@
+# score_ft.py output: ji2025_chili, 33 scored donors
+
+Arms: xcalL_K1000.  Every arm is scored on the same donors against the same observed reference.
+
+## Inputs and fixed contracts
+- Observed reference: `/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ji2025_chili/reference.npz` (sha 19c30f56db69), built from datadir:/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/ji2025_chili;
+  observed cells sum exactly to the features pseudobulk of every donor. Observed labels: /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ji2025_chili/observed_labels/labels/labels.parquet.
+- Generated cells: `generated_counts` read out once with Poisson(seed 20260916) over the **whole NPZ in row order**
+  (`poisson_readout.poisson_integerise`), then the scored donors' rows. These integer counts are used for labels, embedding,
+  depth and the contrast. Pseudobulk uses the summed `generated_counts` (rates), as `score_generated.py`.
+- Labels: frozen f01common scANVI (`predict_scanvi.py`, CPU, zero query epochs), six-type gate ['T cell', 'B cell', 'Monocyte', 'NK cell', 'DC', 'Other']; query genes =
+  the 1,006-gene annotation list minus genes the dataset's sc assay did not measure in every sample (1004 genes).
+- Scoring genes: 6k panel genes measured by both assays in every sample of the dataset (5752 genes).
+
+## Definitions
+1. **Composition**: fraction of each donor's cells per six-type label (all generated cells; all observed QC cells).
+   `comp_l1` = sum over the six labels of |generated - observed| (immicro_common.l1). `comp_l1_obs_loo_null` = L1 of the
+   donor's observed composition to the mean of the *other scored donors'* observed compositions. `composition_top1` =
+   fraction of donors whose generated composition is closest (L1) to their own observed composition.
+2. **Pseudobulk**: `pb_r` = Pearson r of CPM-log1p(summed generated_counts) vs CPM-log1p(observed pseudobulk) on the scoring
+   genes (evaluate_realbulk.cpm_log1p / paired_r). `pb_top1` = the donor's generated pseudobulk is closest (Pearson) to its own
+   observed pseudobulk among the 33 scored donors (immicro_common.retrieval = evaluate_realbulk.top1 arithmetic).
+3. **Precision / coverage / density** (stage_b_embed.py functions): per-cell CP10K over the scoring genes + log1p; PCA-50
+   (randomized, random_state 20260925) and k = 10 real-real k-NN balls fitted on **R1 of the scored donors only**.
+   Reference rule (independent of fold and arm): per donor, cells in canonical order (library_id+barcode), one
+   permutation from `default_rng([20260925, uint64(sha256(donor)[:8])])`; R1 = first n, R2 = next n,
+   n = min(150, n_cells // 2) (4950 R1 cells here). Generated query: per donor n cells (the donor's R1 size),
+   without replacement, `default_rng([20260926, same key])`. Precision = fraction of generated cells inside any R1 ball
+   (Kynkaanniemi 2019); coverage = fraction of R1 cells whose ball holds a generated cell; density = Naeem 2020.
+   Per-donor precision = that donor's generated cells; per-donor coverage = that donor's R1 cells (the pooled values are
+   their n-weighted means; plain means when every donor has 150 reference cells). R2 (held-out real cells of the same donors) is the
+   perfect-generator control: precision 0.9962, coverage 1.0000.
+   **Depth matched**: stage_b_embed depth matching run pairwise (R1 vs this arm): per donor and within-donor rank of detected
+   genes, both cells are thinned by UMI sampling without replacement to the smaller detected-gene count; PCA kept from
+   un-thinned R1, radii recomputed on thinned R1; fresh rng(20260929) per arm. R2 control: precision
+   0.9968, coverage 1.0000.
+4. **Monocyte %** = 100 x Monocyte fraction (all cells) per donor; **median UMI per cell** = median over cells of the
+   Poisson-read (generated) / observed counts summed over the scoring genes (per donor, and pooled).
+5. **Group contrast** (T, B, Monocyte, NK): case groups ['TOL', 'FU1'] (n = 18) vs control groups ['BC', 'AC'] (n = 15); rule: explicit --case-groups / --control-groups; groups from features groups.
+   Genes per type = scoring genes detected (count > 0) in >= 1 % of the scored donors' observed
+   cells of that type. Primary statistic `mean_cp10k`: per-cell CP10K over the scoring genes; per donor the mean over its
+   cells of the type (donors with >= 1 such cell); group mean = mean over the group's donors (donor-balanced);
+   logFC = log2(case mean + 1) - log2(control mean + 1). Secondary `pseudobulk_log1p_cpm`: the SLE benchmark evaluator's
+   statistic (donor x type summed counts, log1p CPM over the scoring genes, mean case - mean control;
+   evaluate_generated_vs_truth.group_log_fold_change). Both are computed on the generated cells (their frozen labels) and on
+   the observed cells; `r_generated` = Pearson r of the two logFC vectors. Broadcast baseline `r_broadcast`: the same
+   observed vector against the logFC of the input bulk (condition NPZ `bulk_counts` of the arm, scoring genes, per-donor
+   CP10K resp. log1p CPM, group means). `partial_r_generated_given_broadcast`: r of the OLS residuals of observed and
+   generated logFC on the bulk logFC (residual_specificity.py).
+6. **Top-50 DEG overlap** (contrast.csv rows `statistic = welch_top50_deg`; summary `deg_top50`): the SLE benchmark's score (`total_top50_overlap` of
+   `bulk2sc-paper/scripts/evaluate_generated_vs_truth.py`, Welch backend: the "top-50 DEG 81 / 84" of the final model), run
+   with the evaluator's own functions on the same donor x type pseudobulks as the contrast. Per cell type, observed and
+   generated separately: donor x type summed counts on all 5752 scoring genes (donors of the two groups with
+   >= 1 cell of the type) -> `welch_bh_deg` (log1p CPM, two-sided Welch t-test case vs control, Benjamini-Hochberg over the
+   genes with a defined p-value; needs >= 2 donors per group) -> `top_deg_genes` (FDR < 0.05, descending |logFC|,
+   first 50; shorter when fewer genes pass). `top50_overlap` = |observed list & generated list|; `top50_jaccard`
+   the evaluator's Jaccard. Summary `sum` = total over the scored types (the quantity reported as 81), `max` = the sum of
+   the observed list lengths (the "of 262"), `sum_4type` = T + B + Monocyte + NK. A type is scored when its observed test
+   is defined; an arm whose generated test is undefined for a scored type gets overlap 0. Types: T cell, B cell, Monocyte, NK cell, DC
+   (aggregates: /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ji2025_chili/deg_reference.npz).
+   **Broadcast analogue** (`top50_overlap_broadcast`, as `sle_broadcast.py` `broadcast_input_bulk`; not part of the
+   published benchmark): the same test on the arm's condition bulk_counts (scoring genes, the 18 + 15
+   donors), one whole-PBMC top-50 list, compared with every type's observed list.
+
+## Files
+- `per_donor.csv`: one row per arm x donor (all per-donor quantities above).
+- `summary.json`: per arm the donor means / pooled values, provenance (NPZ sha, receipt check, generation metadata,
+  Poisson nnz of the whole NPZ, annotation contract), and the reference block (PCA, R2 control).
+- `contrast.csv`: per arm x cell type x statistic (`mean_cp10k`, `pseudobulk_log1p_cpm`: logFC r; `welch_top50_deg`:
+  top-50 DEG overlap, T / B / Monocyte / NK / DC; columns of the other kind are empty).
+- `arms/<arm>/labels/` (predict_scanvi.py labels + manifest), `arms/<arm>/generated_cells.parquet` (NPZ row, donor, label,
+  UMI on the scoring genes, P/C subsample flag), `arms/<arm>/deg_top50_lists.csv` (observed, generated and broadcast
+  top-50 gene lists, Ensembl ids, rank order).