File size: 30,498 Bytes
9892174
 
 
 
 
 
 
 
056d6a5
9892174
 
 
 
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
056d6a5
 
9892174
 
 
056d6a5
6568031
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6568031
 
 
056d6a5
 
9892174
 
 
6568031
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6568031
 
 
056d6a5
 
9892174
 
 
6568031
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6568031
9892174
056d6a5
 
 
 
6568031
 
056d6a5
 
9892174
 
 
 
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
9892174
056d6a5
 
 
 
 
9892174
056d6a5
 
 
9892174
056d6a5
 
 
 
 
 
 
9892174
 
 
 
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b093f70
 
 
 
39e4e52
 
 
 
 
056d6a5
 
 
 
 
 
9892174
056d6a5
 
 
 
 
 
 
 
 
 
9892174
 
 
 
 
 
056d6a5
39e4e52
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
 
056d6a5
 
 
 
 
 
 
9892174
 
003cae0
9892174
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
 
003cae0
056d6a5
 
9892174
 
003cae0
 
9892174
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
 
003cae0
056d6a5
 
9892174
 
003cae0
9892174
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
 
003cae0
056d6a5
 
9892174
 
003cae0
9892174
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
003cae0
056d6a5
 
9892174
 
003cae0
056d6a5
9892174
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
003cae0
056d6a5
 
9892174
 
003cae0
 
9892174
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
003cae0
056d6a5
 
9892174
19934bf
003cae0
 
19934bf
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
19934bf
 
003cae0
056d6a5
 
19934bf
9892174
003cae0
9892174
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
 
003cae0
056d6a5
 
9892174
 
003cae0
9892174
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
003cae0
056d6a5
 
9892174
 
 
 
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
003cae0
 
 
 
 
9892174
056d6a5
 
 
 
 
 
003cae0
056d6a5
 
9892174
 
 
 
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
056d6a5
 
 
 
 
 
 
9892174
 
 
 
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
003cae0
056d6a5
 
9892174
 
 
 
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
9892174
 
 
 
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
 
 
 
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
003cae0
 
 
 
 
 
 
9892174
056d6a5
 
 
 
 
 
 
 
9892174
 
 
 
 
056d6a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9892174
 
056d6a5
 
9892174
 
 
056d6a5
 
9892174
 
056d6a5
9892174
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
{
  "schema_version": "1",
  "generated_at": "2026-05-22",
  "generated_by": "preservation-gap-audit and PACE-to-cloud migration",
  "notes": "Single-source-of-truth inventory of Dolma3 data attribution artifacts across HuggingFace (HCAI-Lab org) and Cloudflare R2 (soc127-dedup bucket). One entry per artifact group. See README.md for navigation by use case.",
  "r2": {
    "endpoint": "https://0934ab8e84ac8f4e81decaf3eb121337.r2.cloudflarestorage.com",
    "bucket": "soc127-dedup",
    "credential_source": "1Password \u2192 Research / Cloudflare R2 API Credentials (scoped tokens issued per consumer team)"
  },
  "artifacts": [
    {
      "id": "dolma3_dedup_corpus",
      "purpose": "Deduplicated 6T-token Dolma3 source corpus (the raw documents)",
      "kind": "r2-prefix",
      "location": {
        "bucket": "soc127-dedup",
        "prefixes": [
          "soc127/phase1_pool_shared/",
          "soc127/phase2_nonpool_final/"
        ]
      },
      "scale": {
        "shards": 58621,
        "approx_tokens": "6T",
        "approx_size_gb": 5000,
        "shard_format": ".jsonl.zst"
      },
      "schema_ref": "docs/DATA_INVENTORY.md \u00a7Source corpus, \u00a7Document format",
      "soc_lineage": [
        "SOC-127"
      ],
      "public_visibility": "private R2 (credentialed read)",
      "consumer_use_case": "Run attribution / training / analysis on raw source documents.",
      "collection": "Source Corpus + Manifest"
    },
    {
      "id": "weborganizer_labels",
      "purpose": "Per-document WebOrganizer topic + format labels (24 topics \u00d7 24 formats, 576 bins)",
      "kind": "hf-dataset-and-r2-prefix",
      "location": [
        {
          "repo_id": "HCAI-Lab/soc91-labels",
          "kind": "dataset",
          "scale": {
            "files": 2719,
            "approx_size_gb": 169.94,
            "format": ".parquet (~60 MB chunks)"
          },
          "note": "HF mirror: 58465 source per-shard parquets concatenated into 2719 ~60MB chunks. Each row carries source_shard_path. README hand-written after script's post-upload write_readme() crash (data itself uploaded cleanly)."
        },
        {
          "bucket": "soc127-dedup",
          "prefixes": [
            "soc91-labels/"
          ],
          "scale": {
            "files": 58465,
            "format": ".parquet (one per source shard) + .stats.json"
          }
        }
      ],
      "schema_ref": "src/dolma/sidecar_manifest_fields.py:43-124 + README \u00a7Sidecar corpus manifest",
      "soc_lineage": [
        "SOC-91",
        "SOC-142 (quality-label fix)"
      ],
      "public_visibility": "private HF dataset + private R2 (both credentialed read)",
      "consumer_use_case": "Filter or stratify documents by topic/format without re-running the classifier. Prefer HF for typical use; R2 prefix preserves per-source-shard granularity if needed.",
      "uploaded_at": "2026-05-23 (HF mirror)",
      "collection": "Source Corpus + Manifest"
    },
    {
      "id": "weborganizer_eda",
      "purpose": "SOC-101 sidecar EDA validation outputs (audit reports, label-distribution stats)",
      "kind": "hf-dataset-and-r2-prefix",
      "location": [
        {
          "repo_id": "HCAI-Lab/soc91-stats",
          "kind": "dataset",
          "note": "Raw R2 mirror of the EDA artifact tree (JSON reports + stats); not Parquet-converted because contents are mostly small JSON files."
        },
        {
          "bucket": "soc127-dedup",
          "prefixes": [
            "soc91-stats/"
          ]
        }
      ],
      "schema_ref": "docs/SOC95_MODAL_RUNBOOK.md \u00a7validation; scripts/modal/sidecar_eda.py",
      "soc_lineage": [
        "SOC-101"
      ],
      "public_visibility": "private HF dataset + private R2",
      "consumer_use_case": "Reference for label-distribution validation; not normally needed by external users.",
      "uploaded_at": "2026-05-23 (HF mirror)",
      "collection": "Source Corpus + Manifest"
    },
    {
      "id": "quality_sidecars",
      "purpose": "Per-doc quality score sidecars (post-SOC-142 label-fix)",
      "kind": "hf-dataset-and-r2-prefix",
      "location": [
        {
          "repo_id": "HCAI-Lab/soc139-quality-sidecars",
          "kind": "dataset",
          "scale": {
            "rows": 1257700000,
            "files": 80,
            "approx_size_gb": 41.8,
            "columns": [
              "doc_id",
              "quality_label_id",
              "quality_score",
              "quality_high_prob",
              "quality_low_prob",
              "quality_confidence",
              "source_shard_path"
            ]
          },
          "note": "HF mirror: per-shard parquets concatenated into 80 chunks (~500MB each)."
        },
        {
          "bucket": "soc127-dedup",
          "prefixes": [
            "soc139-quality-sidecars/"
          ],
          "scale": {
            "files": 58621,
            "format": ".parquet + .label-fix.done + .stats.json"
          }
        }
      ],
      "schema_ref": "src/dolma/quality/sidecar.py (QualityShardStats); src/dolma/sidecar_manifest_fields.py merge logic",
      "soc_lineage": [
        "SOC-139",
        "SOC-142 (inverted-label fix at commit 3342baf)"
      ],
      "public_visibility": "private HF dataset + private R2",
      "consumer_use_case": "Quality-filter the corpus by score threshold. Prefer HF for typical use.",
      "uploaded_at": "2026-05-23 (HF mirror)",
      "collection": "Source Corpus + Manifest"
    },
    {
      "id": "corpus_manifest_dataset",
      "purpose": "Unified per-document manifest joining topic/format + quality + tokens + source shard path. 1.1B docs, 32-column PyArrow schema.",
      "kind": "hf-dataset",
      "location": [
        {
          "repo_id": "HCAI-Lab/dolma3-olmo3-corpus-manifest",
          "kind": "dataset"
        },
        {
          "repo_id": "HCAI-Lab/dolma3-6t-corpus-manifest",
          "kind": "dataset"
        },
        {
          "bucket": "soc127-dedup",
          "prefix": "soc95-manifest/"
        }
      ],
      "scale": {
        "approx_docs": 1098646162,
        "approx_tokens": 2116727590753,
        "approx_size_gb": 213
      },
      "schema_ref": "src/data_attribution/recipes/corpus_manifest.py:15-52",
      "soc_lineage": [
        "SOC-95"
      ],
      "public_visibility": "public HF dataset",
      "consumer_use_case": "One-stop manifest for filtering/stratifying without joining sidecars yourself.",
      "collection": "Source Corpus + Manifest",
      "old_names": [
        "HCAI-Lab/dolma3_6T_corpus_manifest",
        "HCAI-Lab/dolma3_olmo3_corpus_manifest"
      ],
      "renamed_at": "2026-05-25"
    },
    {
      "id": "stratified_samples",
      "purpose": "Materialized stratified working samples (N docs/bin across 576 bins, seed=42)",
      "kind": "hf-multi-dataset",
      "location": [
        {
          "repo_id": "HCAI-Lab/dolma3-6t-sample-500-docs",
          "kind": "dataset",
          "docs": 287936,
          "tokens": 538830013
        },
        {
          "repo_id": "HCAI-Lab/dolma3-6t-sample-1000-docs",
          "kind": "dataset",
          "docs": 575187,
          "tokens": 1083585221
        },
        {
          "repo_id": "HCAI-Lab/dolma3-6t-sample-5000-docs",
          "kind": "dataset",
          "docs": 2855446,
          "tokens": 5299479330
        },
        {
          "repo_id": "HCAI-Lab/dolma3-6t-sample-10000-docs",
          "kind": "dataset",
          "docs": 5678621,
          "tokens": 10506536366
        },
        {
          "repo_id": "HCAI-Lab/dolma3-6t-sample-50000-docs",
          "kind": "dataset",
          "docs": 26249124,
          "tokens": 62819501017
        },
        {
          "repo_id": "HCAI-Lab/dolma3-6t-sample-100000-docs",
          "kind": "dataset",
          "docs": 49709294,
          "tokens": 118399343310,
          "underfilled_bins": 130,
          "twin_bucket": "HCAI-Lab/dolma3-6t-sample-100000-docs",
          "created_at": "2026-05-25"
        }
      ],
      "schema_ref": "docs/WORKING_SAMPLE_DATA_ACCESS.md \u00a7Manifest schema",
      "soc_lineage": [
        "SOC-134"
      ],
      "public_visibility": "public HF datasets",
      "consumer_use_case": "Stratified subsets for training, evaluation, or analysis without materializing your own sample.",
      "collection": "Working Samples + Preconditioner",
      "old_names": [
        "HCAI-Lab/dolma3_6T_sample_10000_docs",
        "HCAI-Lab/dolma3_6T_sample_1000_docs",
        "HCAI-Lab/dolma3_6T_sample_50000_docs",
        "HCAI-Lab/dolma3_6T_sample_5000_docs",
        "HCAI-Lab/dolma3_6T_sample_500_docs"
      ],
      "renamed_at": "2026-05-25"
    },
    {
      "id": "preconditioner_sample",
      "purpose": "100K uniform-random preconditioner sample (no stratification)",
      "kind": "hf-dataset",
      "location": [
        {
          "repo_id": "HCAI-Lab/dolma3-6t-preconditioner-100k",
          "kind": "dataset"
        },
        {
          "repo_id": "HCAI-Lab/dolma3-6t-preconditioner-100k",
          "kind": "bucket"
        }
      ],
      "scale": {
        "docs": 100000,
        "tokens": 251517816,
        "unique_source_shards": 38277,
        "min_tokens_per_doc": 512
      },
      "schema_ref": "docs/WORKING_SAMPLE_DATA_ACCESS.md \u00a7Preconditioner sample",
      "soc_lineage": [
        "SOC-151"
      ],
      "public_visibility": "public HF dataset + public HF bucket",
      "consumer_use_case": "Sample for TrackStar / EK-FAC preconditioner construction.",
      "collection": "Working Samples + Preconditioner",
      "old_names": [
        "HCAI-Lab/dolma3_6T_preconditioner_100k",
        "HCAI-Lab/preconditioner-100k"
      ],
      "renamed_at": "2026-05-25"
    },
    {
      "id": "trackstar_gradient_index_base",
      "purpose": "Per-document gradient index built with Bergson from sample_10000_docs (5.68M docs). 316 shard subdirs, each with gradients.bin, normalizers, preconditioners, configs.",
      "kind": "hf-bucket",
      "location": [
        {
          "repo_id": "HCAI-Lab/trackstar-gradient-index-base",
          "kind": "bucket"
        }
      ],
      "old_names": [
        "HCAI-Lab/tracstar-index-base-soc156"
      ],
      "scale": {
        "shards": 316,
        "approx_size_tb": 1.2,
        "model": "allenai/Olmo-3-1025-7B (base)",
        "projection_dim": 16,
        "precision": "fp32"
      },
      "schema_ref": "docs/TRACKSTAR_DATA_ARTIFACTS.md \u00a75 \u00abGradient Index\u00bb; bergson library docs at github.com/EleutherAI/bergson",
      "soc_lineage": [
        "SOC-156"
      ],
      "public_visibility": "private HF bucket",
      "consumer_use_case": "Score new query sets against the existing training corpus without rebuilding gradients. ~156 GPU-hours to reproduce.",
      "uploaded_at": "2026-05-22",
      "renamed_at": "2026-05-24",
      "collection": "TrackStar Indices + Training Shards"
    },
    {
      "id": "trackstar_training_shards",
      "purpose": "TrackStar training shards with positional doc IDs (id+text only). 316 JSONL files mapping shard_NNNN:INDEX to original Dolma UUID + text.",
      "kind": "hf-dataset",
      "location": [
        {
          "repo_id": "HCAI-Lab/dolma3-6t-sample-10000-docs-trackstar-shards",
          "kind": "dataset"
        }
      ],
      "old_names": [
        "HCAI-Lab/dolma3_6T_sample_10000_docs",
        "HCAI-Lab/dolma3_6T_sample_10000_docs_trackstar_shards",
        "HCAI-Lab/dolma3_6T_sample_10000_docs_tracstar_shards"
      ],
      "scale": {
        "files": 316,
        "approx_size_gb": 41,
        "format": ".jsonl"
      },
      "schema_ref": "docs/TRACKSTAR_DATA_ARTIFACTS.md \u00a73 \u00abDoc ID Mapping\u00bb",
      "soc_lineage": [
        "SOC-156"
      ],
      "public_visibility": "private HF dataset",
      "consumer_use_case": "REQUIRED to resolve positional doc IDs in cloud score matrices back to source text.",
      "uploaded_at": "2026-05-22",
      "renamed_at": "2026-05-25",
      "collection": "TrackStar Indices + Training Shards"
    },
    {
      "id": "trackstar_query_gradients_base",
      "purpose": "Per-query gradient builds for the four OLMES benchmarks (and BBH / ARC for SOC-170 / SOC-171). One Bergson-style index per benchmark dir.",
      "kind": "hf-bucket",
      "location": [
        {
          "repo_id": "HCAI-Lab/trackstar-query-gradients-base",
          "kind": "bucket"
        }
      ],
      "old_names": [
        "HCAI-Lab/tracstar-query-builds-soc156"
      ],
      "scale": {
        "variants": [
          "base",
          "instruct_base",
          "instruct_cot"
        ],
        "approx_size_gb": 17
      },
      "schema_ref": "docs/TRACKSTAR_DATA_ARTIFACTS.md \u00a76 \u00abQuery Gradient Indices\u00bb",
      "soc_lineage": [
        "SOC-156",
        "SOC-170",
        "SOC-171"
      ],
      "public_visibility": "private HF bucket",
      "consumer_use_case": "Optional. Needed if scoring additional training data against existing query sets.",
      "uploaded_at": "2026-05-22",
      "renamed_at": "2026-05-24",
      "collection": "TrackStar Indices + Training Shards"
    },
    {
      "id": "trackstar_preconditioners",
      "purpose": "TrackStar mixed preconditioners for OLMo3 base, instruct, and think model variants.",
      "kind": "hf-bucket",
      "location": [
        {
          "repo_id": "HCAI-Lab/trackstar-preconditioners",
          "kind": "bucket"
        }
      ],
      "old_names": [
        "HCAI-Lab/tracstar-preconditioners"
      ],
      "scale": {
        "models": 3,
        "approx_size_mb": 885,
        "files": 78
      },
      "soc_lineage": [
        "SOC-152",
        "SOC-168"
      ],
      "public_visibility": "public HF bucket",
      "consumer_use_case": "Apply pre-built preconditioner to new attribution scoring runs. NOTE: preconditioner must match the model used for gradient build (SOC-162 finding).",
      "renamed_at": "2026-05-24",
      "collection": "TrackStar Indices + Training Shards"
    },
    {
      "id": "trackstar_scores_base_olmes_4bench",
      "purpose": "Per-query attribution score matrices for the OLMo-3-7B base run, 4 OLMES benchmarks (gsm8k, mmlu_social_science, mmlu_stem, socialiqa) \u00d7 316 shards = 2532 .npy files.",
      "kind": "hf-bucket",
      "location": [
        {
          "repo_id": "HCAI-Lab/trackstar-scores-base-olmes-4bench",
          "kind": "bucket"
        }
      ],
      "old_names": [
        "HCAI-Lab/tracstar-scores-base-soc156"
      ],
      "scale": {
        "benchmarks": [
          "gsm8k",
          "mmlu_social_science",
          "mmlu_stem",
          "socialiqa"
        ],
        "files": 2532,
        "approx_size_gb": 396,
        "model": "allenai/Olmo-3-1025-7B (base)"
      },
      "schema_ref": "docs/TRACKSTAR_DATA_ARTIFACTS.md \u00a71 \u00abFull Score Matrices\u00bb",
      "soc_lineage": [
        "SOC-156"
      ],
      "public_visibility": "public HF bucket",
      "consumer_use_case": "Influence(doc_i, query_j) lookup for the four OLMES benchmarks on OLMo-3-7B base.",
      "renamed_at": "2026-05-24",
      "collection": "TrackStar Scores + Analysis"
    },
    {
      "id": "trackstar_scores_instruct_cot_olmes_4bench",
      "purpose": "Score matrices for the OLMo-3-7B instruct-cot (think) run, 4 OLMES benchmarks (same shape as base).",
      "kind": "hf-bucket",
      "location": [
        {
          "repo_id": "HCAI-Lab/trackstar-scores-instruct-cot-olmes-4bench",
          "kind": "bucket"
        }
      ],
      "old_names": [
        "HCAI-Lab/tracstar-scores-instruct-cot-soc161"
      ],
      "scale": {
        "files": 2532,
        "approx_size_gb": 396,
        "model": "allenai/Olmo-3-1025-7B-think"
      },
      "soc_lineage": [
        "SOC-161"
      ],
      "public_visibility": "public HF bucket",
      "consumer_use_case": "Same as base scores but for the instruct-cot model variant.",
      "renamed_at": "2026-05-24",
      "collection": "TrackStar Scores + Analysis"
    },
    {
      "id": "trackstar_scores_instruct_cot_noprecond_mmlu_ss",
      "purpose": "Score matrices for the OLMo-3-7B instruct-cot run with NO preconditioner applied (single benchmark: MMLU social science only). Ablation comparison to the preconditioned scores in the parent OLMES-4bench bucket.",
      "kind": "hf-bucket",
      "location": [
        {
          "repo_id": "HCAI-Lab/trackstar-scores-instruct-cot-noprecond-mmlu-ss",
          "kind": "bucket"
        }
      ],
      "old_names": [
        "HCAI-Lab/tracstar-scores-instruct-cot-noprecond-soc161"
      ],
      "scale": {
        "benchmarks": [
          "mmlu_social_science"
        ],
        "files": 633,
        "approx_size_gb": 70.0,
        "model": "allenai/Olmo-3-1025-7B-think",
        "note": "316 shard_NNNN.npy (221 MB each) + 316 shard_NNNN_doc_ids.json + 1 query_ids.json"
      },
      "schema_ref": "docs/TRACKSTAR_DATA_ARTIFACTS.md \u00a71 \u00abFull Score Matrices\u00bb",
      "soc_lineage": [
        "SOC-161 (noprecond ablation variant)"
      ],
      "public_visibility": "public HF bucket",
      "consumer_use_case": "Compare preconditioned vs raw influence scores for the same model+benchmark to study preconditioner contribution.",
      "uploaded_at": "2026-05-23",
      "renamed_at": "2026-05-24",
      "collection": "TrackStar Scores + Analysis"
    },
    {
      "id": "trackstar_scores_bbh",
      "purpose": "BBH attribution scores for base and instruct_base model variants.",
      "kind": "hf-bucket",
      "location": [
        {
          "repo_id": "HCAI-Lab/trackstar-scores-base-bbh",
          "kind": "bucket"
        },
        {
          "repo_id": "HCAI-Lab/trackstar-scores-instruct-base-bbh",
          "kind": "bucket"
        }
      ],
      "old_names": [
        "HCAI-Lab/tracstar-scores-base-soc170",
        "HCAI-Lab/tracstar-scores-instruct_base-soc170"
      ],
      "scale": {
        "files_each": 1899,
        "approx_size_gb_each": 14
      },
      "soc_lineage": [
        "SOC-170"
      ],
      "public_visibility": "public HF buckets",
      "consumer_use_case": "BBH-task attribution analysis.",
      "renamed_at": "2026-05-24",
      "collection": "TrackStar Scores + Analysis"
    },
    {
      "id": "trackstar_scores_gsm8k_arc",
      "purpose": "GSM8K + ARC attribution scores for base and instruct_base.",
      "kind": "hf-bucket",
      "location": [
        {
          "repo_id": "HCAI-Lab/trackstar-scores-base-gsm8k-arc",
          "kind": "bucket"
        },
        {
          "repo_id": "HCAI-Lab/trackstar-scores-instruct-base-gsm8k-arc",
          "kind": "bucket"
        }
      ],
      "old_names": [
        "HCAI-Lab/tracstar-scores-base-soc171",
        "HCAI-Lab/tracstar-scores-instruct_base-soc171"
      ],
      "scale": {
        "files_each": 1899,
        "approx_size_gb_each": 79
      },
      "soc_lineage": [
        "SOC-171"
      ],
      "public_visibility": "public HF buckets",
      "renamed_at": "2026-05-24",
      "collection": "TrackStar Scores + Analysis"
    },
    {
      "id": "top_k_results",
      "purpose": "Ranked top-K influence results (top-100 JSONL and top-2K parquet/CSV) per benchmark.",
      "kind": "hf-mixed",
      "location": [
        {
          "repo_id": "HCAI-Lab/dolma3-trackstar-influence-scores",
          "kind": "dataset",
          "private": true,
          "files": [
            "influence_scores_full.parquet",
            "top2k_{gsm8k,mmlu_socsci,mmlu_stem,socialiqa}.{csv,parquet}"
          ]
        },
        {
          "repo_id": "HCAI-Lab/trackstar-top2k-base-gsm8k-arc",
          "kind": "bucket",
          "soc": "SOC-171"
        },
        {
          "repo_id": "HCAI-Lab/trackstar-top2k-instruct-base-gsm8k-arc",
          "kind": "bucket",
          "soc": "SOC-171"
        }
      ],
      "old_names": [
        "HCAI-Lab/dolma3-tracstar-influence-scores",
        "HCAI-Lab/tracstar-top2k-soc171-base",
        "HCAI-Lab/tracstar-top2k-soc171-instruct-base"
      ],
      "schema_ref": "docs/TRACKSTAR_DATA_ARTIFACTS.md \u00a72 \u00abTop-100 Ranked Results\u00bb",
      "soc_lineage": [
        "SOC-156",
        "SOC-164",
        "SOC-171"
      ],
      "consumer_use_case": "Drop-in ranked influence results for the four canonical OLMES benchmarks; cheapest path for analysis that only needs top-K docs per query.",
      "renamed_at": "2026-05-24",
      "collection": "TrackStar Scores + Analysis"
    },
    {
      "id": "query_data",
      "purpose": "OLMES evaluation queries used for attribution (prompt + answer + correctness label).",
      "kind": "hf-multi-dataset",
      "location": [
        {
          "repo_id": "HCAI-Lab/base-query-data",
          "kind": "dataset",
          "model": "OLMo-3-7B base"
        },
        {
          "repo_id": "HCAI-Lab/instruct-query-data",
          "kind": "dataset",
          "model": "OLMo-3-7B instruct"
        },
        {
          "repo_id": "HCAI-Lab/instruct-cot-query-data",
          "kind": "dataset",
          "model": "OLMo-3-7B instruct-cot"
        },
        {
          "repo_id": "HCAI-Lab/archive-reduced-queries",
          "kind": "dataset"
        },
        {
          "repo_id": "HCAI-Lab/archive-common-sense-queries",
          "kind": "dataset"
        }
      ],
      "scale": {
        "benchmarks": [
          "gsm8k:1319",
          "mmlu_social_science:3077",
          "mmlu_stem:3018",
          "socialiqa:10000"
        ]
      },
      "schema_ref": "docs/TRACKSTAR_DATA_ARTIFACTS.md \u00a74 \u00abQuery ID Mapping\u00bb",
      "soc_lineage": [
        "SOC-119",
        "SOC-155 (LLM-verified SocialIQA reasoning types)"
      ],
      "public_visibility": "public HF datasets",
      "consumer_use_case": "Reuse the same query sets for new attribution runs to keep results comparable.",
      "collection": "Query Data",
      "old_names": [
        "HCAI-Lab/common-sense-queries",
        "HCAI-Lab/reduced-queries"
      ],
      "renamed_at": "2026-05-25"
    },
    {
      "id": "olmes_eval_results",
      "purpose": "OLMES benchmark evaluation results across model variants.",
      "kind": "hf-multi-dataset",
      "location": [
        {
          "repo_id": "HCAI-Lab/olmes-eval-olmo3-7b-base",
          "kind": "dataset"
        },
        {
          "repo_id": "HCAI-Lab/olmes-eval-olmo3-7b-instruct-base",
          "kind": "dataset"
        },
        {
          "repo_id": "HCAI-Lab/olmes-eval-olmo3-7b-instruct-cot",
          "kind": "dataset"
        },
        {
          "repo_id": "HCAI-Lab/olmes-eval-olmo3-7b-thinking",
          "kind": "dataset"
        },
        {
          "repo_id": "HCAI-Lab/olmes-eval-olmo3-7b-think-mc",
          "kind": "dataset",
          "private": true
        },
        {
          "repo_id": "HCAI-Lab/olmes-eval-olmo3-7b-think-cot",
          "kind": "dataset",
          "private": true
        },
        {
          "repo_id": "HCAI-Lab/olmes-eval-smollm3-3b-base",
          "kind": "dataset"
        },
        {
          "repo_id": "HCAI-Lab/trackstar-olmes-eval-artifacts",
          "kind": "bucket",
          "soc": "SOC-166"
        }
      ],
      "old_names": [
        "HCAI-Lab/data-attribution-olmo-3-7B-base-evaluation",
        "HCAI-Lab/data-attribution-olmo-3-7B-instruct-base-evaluation",
        "HCAI-Lab/data-attribution-olmo-3-7B-instruct-cot-evaluation",
        "HCAI-Lab/data-attribution-olmo-3-7B-think-cot-evaluation",
        "HCAI-Lab/data-attribution-olmo-3-7B-think-mc-evaluation",
        "HCAI-Lab/data-attribution-olmo-3-7B-thinking-evaluation",
        "HCAI-Lab/data-attribution-smollm3-3b-base-evaluation",
        "HCAI-Lab/soc166-olmes-eval-artifacts"
      ],
      "soc_lineage": [
        "SOC-166",
        "SOC-167"
      ],
      "consumer_use_case": "Inspect per-query model predictions and correctness used in attribution analysis.",
      "renamed_at": "2026-05-25",
      "collection": "OLMES Evaluations"
    },
    {
      "id": "dedup_state",
      "purpose": "Deduplication artifacts: Bloom filter, doc IDs, unique docs materialization.",
      "kind": "hf-multi-dataset",
      "location": [
        {
          "repo_id": "HCAI-Lab/dolma3-6t-bloom-index",
          "kind": "dataset",
          "purpose": "Bloom filter (~unique IDs)"
        },
        {
          "repo_id": "HCAI-Lab/archive-dolma3-6t-doc-ids-pershard",
          "kind": "dataset",
          "purpose": "Per-shard doc ID lists (gzip JSONL)",
          "private": true
        },
        {
          "repo_id": "HCAI-Lab/archive-dolma3-6t-mix-doc-ids",
          "kind": "dataset",
          "purpose": "Mix doc IDs",
          "private": true
        },
        {
          "repo_id": "HCAI-Lab/dolma3-6t-unique",
          "kind": "dataset",
          "purpose": "1.258B unique docs materialized"
        },
        {
          "repo_id": "HCAI-Lab/archive-dolma3-6t-dedup-state",
          "kind": "dataset",
          "purpose": "Bloom + dedup backup",
          "private": true
        }
      ],
      "schema_ref": "src/dolma/provenance.py",
      "soc_lineage": [
        "SOC-90",
        "SOC-127"
      ],
      "consumer_use_case": "Reference for dedup decisions; reproduce the dedup pipeline.",
      "collection": "Archive (pre-6T and legacy)",
      "old_names": [
        "HCAI-Lab/dolma3_6t_doc_ids",
        "HCAI-Lab/dolma3_6t_mix_doc_ids",
        "HCAI-Lab/dolma3_6t_unique",
        "HCAI-Lab/soc-90-keep-backup"
      ],
      "renamed_at": "2026-05-25"
    },
    {
      "id": "earlier_enrichment_pool",
      "purpose": "Pre-6T enrichment work: 150B pool sample plus enriched/cleaned variants.",
      "kind": "hf-multi-dataset",
      "location": [
        {
          "repo_id": "HCAI-Lab/archive-dolma3-pool-150b",
          "kind": "dataset"
        },
        {
          "repo_id": "HCAI-Lab/archive-dolma3-pool-150b-cleaned",
          "kind": "dataset"
        },
        {
          "repo_id": "HCAI-Lab/archive-dolma3-pool-150b-enriched",
          "kind": "dataset"
        },
        {
          "repo_id": "HCAI-Lab/archive-dolma3-pool-150b-samples",
          "kind": "dataset",
          "private": true
        },
        {
          "repo_id": "HCAI-Lab/archive-dolma3-pool-stratified",
          "kind": "dataset"
        },
        {
          "repo_id": "HCAI-Lab/archive-dolma3-mix-150b-enriched",
          "kind": "dataset",
          "private": true
        },
        {
          "repo_id": "HCAI-Lab/archive-dolma3-enriched-100k",
          "kind": "dataset"
        },
        {
          "repo_id": "HCAI-Lab/archive-dolma3-enriched-small",
          "kind": "dataset"
        }
      ],
      "soc_lineage": [
        "SOC-5",
        "SOC-7"
      ],
      "consumer_use_case": "Reference for earlier enrichment / sampling work. Not needed for current attribution use.",
      "collection": "Archive (pre-6T and legacy)",
      "old_names": [
        "HCAI-Lab/dolma3-enriched-100k",
        "HCAI-Lab/dolma3_mix_150B_enriched",
        "HCAI-Lab/dolma3_pool_150B",
        "HCAI-Lab/dolma3_pool_150B_cleaned",
        "HCAI-Lab/dolma3_pool_150B_enriched",
        "HCAI-Lab/dolma3_pool_150B_samples",
        "HCAI-Lab/dolma3_pool_stratified",
        "HCAI-Lab/dolma_enriched_small"
      ],
      "renamed_at": "2026-05-25"
    },
    {
      "id": "visualization_artifacts",
      "purpose": "Figures and analysis outputs (heatmaps, bin-characterization, paper figures).",
      "kind": "hf-multi-bucket",
      "location": [
        {
          "repo_id": "HCAI-Lab/dolma3-bin-characterization",
          "kind": "bucket"
        },
        {
          "repo_id": "HCAI-Lab/trackstar-bbh-eval-artifacts",
          "kind": "bucket"
        },
        {
          "repo_id": "HCAI-Lab/trackstar-figures-gsm8k-arc",
          "kind": "bucket"
        },
        {
          "repo_id": "HCAI-Lab/trackstar-analysis-gsm8k-arc",
          "kind": "bucket"
        },
        {
          "repo_id": "HCAI-Lab/dolma3-influence-heatmaps",
          "kind": "dataset"
        }
      ],
      "old_names": [
        "HCAI-Lab/soc14-rq4-bin-characterization",
        "HCAI-Lab/soc170-bbh-artifacts",
        "HCAI-Lab/tracstar-figures-soc171",
        "HCAI-Lab/tracstar-analysis-soc171",
        "HCAI-Lab/soc28-influence-heatmaps"
      ],
      "soc_lineage": [
        "SOC-14",
        "SOC-28",
        "SOC-170",
        "SOC-171"
      ],
      "renamed_at": "2026-05-24",
      "collection": "TrackStar Scores + Analysis"
    },
    {
      "id": "job_archive",
      "purpose": "Archival snapshot of PACE SLURM jobs, logs, dedup state, and small ancillary work outputs.",
      "kind": "hf-dataset",
      "location": [
        {
          "repo_id": "HCAI-Lab/dolma3-attribution-job-archive",
          "kind": "dataset"
        }
      ],
      "soc_lineage": [
        "SOC-14",
        "SOC-90",
        "SOC-149",
        "SOC-159",
        "SOC-161",
        "SOC-167",
        "SOC-171"
      ],
      "public_visibility": "private HF dataset",
      "consumer_use_case": "Reproducibility audits. Not normally needed.",
      "uploaded_at": "2026-05-22",
      "collection": "Archive (pre-6T and legacy)"
    }
  ],
  "schema_references": {
    "source_shard_jsonl": "docs/DATA_INVENTORY.md \u00a7Document format",
    "working_sample_manifest": "docs/WORKING_SAMPLE_DATA_ACCESS.md \u00a7Manifest schema (6 columns: doc_id, token_count, shard_path, bin_id, bin_topic, bin_format)",
    "corpus_manifest_32col": "src/data_attribution/recipes/corpus_manifest.py lines 15-52",
    "sidecar_row_construction": "src/dolma/sidecar_manifest_fields.py lines 43-124",
    "score_matrix_format": "docs/TRACKSTAR_DATA_ARTIFACTS.md \u00a71",
    "bin_id_formula": "topic_idx * len(FORMATS) + format_idx + 1 (576 bins, 1-indexed); src/dolma/manifest_fields.py:60-67"
  },
  "code_repo": "https://github.com/eilab-gt/social-data-attribution",
  "credentials": {
    "hf": "Token at ~/.hf_token on PACE; or HF_TOKEN env var. HCAI-Lab admin (Glenn) grants collaborator access for private repos.",
    "r2": "Read-only API token via 1Password Share. R2_ACCESS_KEY_ID + R2_SECRET_ACCESS_KEY env vars. scripts/bootstrap/with_r2_credentials.sh wraps the lookup.",
    "pace": "Not granted to external consumers. PACE access requires Georgia Tech sponsorship."
  }
}