Cross-lingual Code Generation Performance via Intermediate English Task Training

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: How does intermediate-task training on English code-related tasks (e.g., CodeNet, HumanEval) affect zero-shot cross-lingual performance on code generation benchmarks like MBPP-X and X-CoderEval compared to models directly fine-tuned on multilingual code datasets?

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