Language-Conditioned Token and Reasoning Efficiency in Large Language Models: A Paired Cross-Lingual Study Protocol

Excerpt

arXiv:2610.04295v1 Announce Type: new Abstract: Large language models incur language-dependent representation and inference costs, but existing comparisons often conflate input language, assigned observable-trace language, and answer realization. We specify a prospective paired study that separates these interfaces while holding the semantic item, checkpoint, and answer oracle fixed. The initial design instantiates 240 exactly scored items rendered from templates in English and seven non-English