New Research from Welo Data Uses Bilingual Prompts to Uncover Performance Differences in LLM Causal Reasoning