arxivApr 21
Macaron: Controlled, Human-Written Benchmark for Multilingual and Multicultural Reasoning via Template-Filling
arXiv:2602.10732v2 Announce Type: replace Abstract: Multilingual benchmarks rarely test reasoning over culturally grounded premises: translated datasets keep English-centric scenarios, while culture-first datasets often lack control over the reasoning required. We propose Macaron, a template-first b