Multilingual Knowledge Distillation for Zero-Shot Cross-Lingual Retrieval in Low-Resource Languages

Recent advances in training multilingual language models on large datasets seem to have shown promising results in knowledge transfer across languages and achieve high performance on downstream tasks. However, we question to what extent the current evaluation benchmarks and setups accurately measure zero-shot cross-lingual knowledge transfer. In this work, we challenge the assumption that high zero-shot performance on target tasks reflects high cross-lingual ability by introducing more challenging setups involving instances with multiple languages. Through extensive experiments and analysis, w

Research goal: How does multilingual knowledge distillation from high-resource languages impact the zero-shot cross-lingual retrieval performance of low-resource languages when evaluated on the XNLI benchmark compared to the MLQA benchmark?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.1/10.