This paper studies zero-shot cross-lingual transfer of vision-language models. Specifically, we focus on multilingual text-to-video search and propose a Transformer-based model that learns contextualized multilingual multimodal embeddings. Under a zero-shot setting, we empirically demonstrate that performance degrades significantly when we query the multilingual text-video model with non-English sentences. To address this problem, we introduce a multilingual multimodal pre-training strategy, and collect a new multilingual instructional video dataset (MultiHowTo100M) for pre-training. ExperimenResearch goal: How does the performance of multimodal vision-language models (e.g., Flamingo) on zero-shot cross-lingual MMLU compare to text-only models when fine-tuned on multilingual intermediate tasks?Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.2/10.