Multimodal Model Robustness in Zero-Shot Cross-Lingual Retrieval via Code-Switched Fine-Tuning

Transferring information retrieval (IR) models from a high-resource language (typically English) to other languages in a zero-shot fashion has become a widely adopted approach. In this work, we show that the effectiveness of zero-shot rankers diminishes when queries and documents are present in different languages. Motivated by this, we propose to train ranking models on artificially code-switched data instead, which we generate by utilizing bilingual lexicons. To this end, we experiment with lexicons induced from (1) cross-lingual word embeddings and (2) parallel Wikipedia page titles. We use

Research goal: Do multimodal models (e.g., CLIP, BLIP) exhibit improved robustness in zero-shot cross-lingual image-text retrieval when fine-tuned on artificially code-switched captions, as measured by R@K scores compared to standard monolingual training?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.3/10.