Good first issue: verify exact CLI versions and budgets in one paper dossier #11

Pick one included paper whose product version, run count, or budget is marked not-reported or unknown, then verify it against the official full text, appendix, supplement, and author-maintained artifact.

Checklist

• Preserve the exact model snapshot string; do not shorten or normalize it.
• Record the product or CLI version only when the source states it.
• Check reasoning mode, temperature, output tokens, time or turn limits, tool permissions, and number of runs.
• Add the exact section, table, appendix, or artifact path to evidence.source_location.
• Keep unknown when the evidence is genuinely absent; do not infer from release dates or model knowledge.
• Update data/papers.yaml, run make build, make check, and make site-check.

Start from papers/README.md and CONTRIBUTING.md. Comment with the catalog ID you plan to audit so work is not duplicated.