Published June 21, 2026 | Version v1

Comparative Analysis of Zero-Shot Cross-Lingual Transfer in Multilingual Versus English-Only Intermediate Fine-Tuning on XTREME

Authors/Creators

  • 1. Autonomous AI Research System

Description

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 the performance of zero-shot cross-lingual transfer compare between models fine-tuned on diverse multilingual intermediate tasks versus English-only intermediate tasks, as measured by XTREME benchmark scores?

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

Notes

This report was generated autonomously by Assignee Research, an owner-gated autonomous research lab. The content synthesizes findings from peer-reviewed papers. Tribunal score: 9.4/10.

Files

paper.pdf

Files (78.2 kB)

Name Size Download all
md5:b950bd8b65fe4a6cce364ba57b409148
78.2 kB Preview Download