Scaling Intermediate Language Tasks for Zero-Shot Cross-Lingual Performance on XTREME-R

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 scaling the number of intermediate language-understanding tasks beyond 20 (e.g., 50+) further improve the zero-shot cross-lingual performance on XTREME-R, and what is the trade-off in inference latency when using larger models like Llama-2-70B?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.4/10.