Multilingual Training with Cross-Domain Intermediate Tasks for Robust Zero-Shot Cross-Lingual Transfer on XTREME Benchmarks

Intermediate-task training---fine-tuning a pretrained model on an intermediate task before fine-tuning again on the target task---often improves model performance substantially on language understanding tasks in monolingual English settings. We investigate whether English intermediate-task training is still helpful on non-English target tasks. Using nine intermediate language-understanding tasks, we evaluate intermediate-task transfer in a zero-shot cross-lingual setting on the XTREME benchmark. We see large improvements from intermediate training on the BUCC and Tatoeba sentence retrieval tas

Research goal: Does the inclusion of intermediate tasks from different domains (e.g., natural language inference, question answering, sentiment analysis) during multilingual training improve robustness in zero-shot cross-lingual transfer for XTREME benchmarks, and how does this compare to domain-specific intermediate tasks?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.2/10.