ChainQueen's citation practice splits three ways: formal citation for its claimed mathematical lineage and its own prior software, none for third-party infrastructure like TensorFlow

chainqueentaichimaterial-point-methodtensorflowsoftware-citationcitation-historyhistory-of-computingyuanming-hu

ChainQueen's 34-item reference list treats three categories of antecedent differently. (1) The mathematical/physical method it extends — the Material Point Method — is credited to a formal academic citation, [[claim-chainqueen-reference-list-directly-cites-sulsky-zhou-schreyer-1995-mpm-origin|reference [9], Sulsky, Zhou & Schreyer (1995)]]. (2) The paper's own underlying software library — Taichi, written by ChainQueen's own first author, Yuanming Hu — also receives a formal citation to its own arXiv paper, introduced in-text as "‡ Based the Taichi [30] open source computer graphics library" (the paper's own typo, printed as-is). (3) TensorFlow, the automatic-differentiation framework the paper's reference implementation runs on and is explicitly benchmarked against ("TF indicates the TensorFlow implementation"), receives no bibliography entry anywhere among the 34 references — confirmed by an end-to-end read of the full list.

The split is not "software is never cited" (Taichi disproves that) and not "only mathematics is cited" (Taichi disproves that too). It tracks whether the antecedent is claimed as the paper's own intellectual lineage — the MPM method it extends, its own prior-authored tool — versus used as an uncredited third-party utility. TensorFlow is invoked the way a lab invokes an instrument, not the way a paper invokes a predecessor.