mmT5: Modular Multilingual Pre-Training Solves Source Language Hallucinations
This addresses source language hallucinations in multilingual models, improving performance for tasks across 40+ languages, though it is incremental as it builds on existing modular approaches.
The paper tackled the problem of multilingual sequence-to-sequence models performing poorly with increased language coverage and failing to generate text in the correct target language in few-shot settings, resulting in mmT5 outperforming mT5 by raising the rate of generating text in the correct language under zero-shot settings from 7% to 99%.
Multilingual sequence-to-sequence models perform poorly with increased language coverage and fail to consistently generate text in the correct target language in few-shot settings. To address these challenges, we propose mmT5, a modular multilingual sequence-to-sequence model. mmT5 utilizes language-specific modules during pre-training, which disentangle language-specific information from language-agnostic information. We identify representation drift during fine-tuning as a key limitation of modular generative models and develop strategies that enable effective zero-shot transfer. Our model outperforms mT5 at the same parameter sizes by a large margin on representative natural language understanding and generation tasks in 40+ languages. Compared to mT5, mmT5 raises the rate of generating text in the correct language under zero-shot settings from 7% to 99%, thereby greatly alleviating the source language hallucination problem.