Mitigating Performance Gaps in Cross-Lingual Retrieval via Optimal Transport Distillation and Adversarial Training

Benefiting from transformer-based pre-trained language models, neural ranking models have made significant progress. More recently, the advent of multilingual pre-trained language models provides great support for designing neural cross-lingual retrieval models. However, due to unbalanced pre-training data in different languages, multilingual language models have already shown a performance gap between high and low-resource languages in many downstream tasks. And cross-lingual retrieval models built on such pre-trained models can inherit language bias, leading to suboptimal result for low-reso

Research goal: Can the gap in performance between high- and low-resource languages in cross-lingual retrieval be mitigated by combining optimal transport distillation with adversarial training, as measured by accuracy and cross-lingual transferability on benchmark datasets such as XQuAD or PAWS-X?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.3/10.