Financial institutions are increasingly deploying large language models (LLMs) in multilingual markets, raising a practical question: can pipelines built and validated in English be reused across languages, or must they be rebuilt for each locale? This study evaluates cross‑lingual reusability along three dimensions—accuracy, reasoning‑effort transfer, and behavioral stability—through three complementary tasks.
Task 1: cheque extraction. Task 2: Mintaka QA with pinned evidence, covering nine languages and nine model‑effort configurations, testing whether accuracy and effort settings transfer. Task 3: FinanceBench pilot, selecting 50 items from SEC 10‑K filings in four languages, validating whether the observed patterns generalize to financial content.
The findings confirm broad accuracy reusability. Gemini 3.5 Flash achieves 95.2%‑97.6% extraction accuracy across prompt languages, and FinanceBench per‑language means differ by only 5.8 percentage points (77.1%‑82.9%). Notably, the language‑difficulty ranking reverses between domains: Hindi performs worst on Mintaka but best on FinanceBench at 82.9%.
Effort reusability is domain‑dependent: higher reasoning effort yields marginal gains on general‑knowledge QA but up to 8 percentage points on financial content; effort configurations that help in English proportionally help in other languages as well.
Behavioral stability holds: citation‑set overlap with English averages 77%‑79% Jaccard, with no single language consistently worse.
Overall, English‑built pipelines can be qualified for reuse, yet per‑domain and per‑language validation remains necessary—semantic similarity or self‑reported confidence do not reliably predict cross‑lingual accuracy.
Review: The paper offers a systematic cross‑lingual reuse evaluation framework, demonstrating transferable performance for financial tasks while cautioning practitioners to perform fine‑grained language‑and‑domain checks before deployment.