Effect of Optimal Transport Distillation on Cross-Lingual Retrieval Scaling in Low-Resource Languages

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: What is the effect of optimal transport distillation on the scaling behavior of cross-lingual retrieval models measured by nDCG@10 improvements on the MIRACL benchmark as the number of low-resource training languages increases?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.1/10.