Link the arXiv paper (2609.03350)

Card: explicit train split config

Re-chunk parquet (10K-row groups, page index) for the dataset viewer

Sync with paper v2: subset manifest, comparable corpus sizes

Card: token counts, AceReason answer-only disclosure, audit wording

CPT training-subset manifest (104,630 ids present verbatim in the release)

Author order update

Full author list + team section

Update card links to COPA-AI

ArmSTEM v1: 372,907 verified parallel EN-HY pairs + stats

Card: solver_limited field, stats dir

Add release card

initial commit