Dolma3 — Source Corpus + Manifest - a HCAI-Lab Collection

HCAI-Lab 's Collections

updated May 25

Dolma3 6T source corpus, dedup state, and unified per-doc manifests for the attribution project.

HCAI-Lab/dolma3-olmo3-corpus-manifest

Viewer • Updated May 25 • 1.34M • 131

Note Unified per-doc manifest (1.1B docs, 32-col schema with topic+format+quality+token+shard).
HCAI-Lab/dolma3-6t-corpus-manifest

Preview • Updated May 25 • 453

Note Alternative-shape unified corpus manifest.
HCAI-Lab/dolma3-6t-bloom-index

Updated Mar 13 • 24

Note Bloom filter encoding the deduplicated doc-id set from SOC-90.
HCAI-Lab/dolma3-6t-unique

Preview • Updated May 25 • 467 • 1

Note 1.258B materialized unique docs after dedup.
HCAI-Lab/soc91-labels

Viewer • Updated May 23 • 1.16B • 1.29k

Note Per-doc WebOrganizer topic+format labels (R2-mirror).
HCAI-Lab/soc91-stats

Updated May 23 • 385

Note SOC-101 sidecar EDA validation outputs (R2-mirror).
HCAI-Lab/soc139-quality-sidecars

Viewer • Updated May 23 • 1.26B • 93

Note Per-doc quality classifier scores (R2-mirror, post-SOC-142 fix).