repro-csr / PROVENANCE.md
SabaPivot's picture
Audit Claim 1 dataset scope from hash-pinned author source
38ba967 verified
|
Raw
History Blame Contribute Delete
1.59 kB

Provenance for the 2026-08-03 source audits

Claim 1 scope audit

  • Paper: arXiv 2601.19210, latest source downloaded 2026-08-03
  • Source archive SHA-256: 47d47f10240a7bc5bcd370a4eac0639c3343c4caa4ceeb844db2a34c23612977
  • Audited source location: Sections/6_Exp.tex, Table tab:comparison_strong
  • Audit script: evidence/claim1_scope_audit/csr_claim1_scope_audit.py
  • Audit output: evidence/claim1_scope_audit/claim1_scope_audit.json
  • Script SHA-256: b7d26b72e1e3bf5f285008bfd767cf428e8a446bbda1a109d8578ec0921141f5
  • Output SHA-256: e3b7e4dd5e2dfc487e958aa81a3928d82f633a13f521d48feea0bf767c13ef69
  • Arithmetic: exact decimal, group counts 6/4/2/4
  • Hardware: CPU only; no GPU used

This audit checks whether 66.4 is the all-16-dataset mean stated in the registered claim. It does not dispute the narrower General-group result.

Claim 5 wording audit

  • Paper: arXiv 2601.19210, latest source downloaded 2026-08-03
  • Source archive SHA-256: 47d47f10240a7bc5bcd370a4eac0639c3343c4caa4ceeb844db2a34c23612977
  • Audited source location: Sections/6_Exp.tex, Table tab:results
  • Audit script: evidence/table7_scope_audit/csr_table7_audit.py
  • Audit output: evidence/table7_scope_audit/table7_audit.json
  • Script SHA-256: ba98204a4b5dbcde80736a7c0da426d8dea316203d1c1514bd58cb74c3d312df
  • Output SHA-256: 168d2b8daf9c778eae7e4381670323579d164ff75b46d867acaeba592fb470c1
  • Arithmetic: exact decimal subtraction
  • Hardware: CPU only; no GPU used

This audit checks what Table 7's numbers mean. It does not rerun the segmentation, captioning, or VQA pipelines.