Impact of Intermediate Task Scaling on Zero-Shot Cross-Lingual Transfer in XTREME-R

Transfer learning from large language models (LLMs) has emerged as a powerful technique to enable knowledge-based fine-tuning for a number of tasks, adaptation of models for different domains and even languages. However, it remains an open question, if and when transfer learning will work, i.e. leading to positive or negative transfer. In this paper, we analyze the knowledge transfer across three natural language processing (NLP) tasks - text classification, sentimental analysis, and sentence similarity, using three LLMs - BERT, RoBERTa, and XLNet - and analyzing their performance, by fine-tun

Research goal: Does increasing the scale of intermediate task training (e.g., larger datasets or more tasks) improve zero-shot cross-lingual transfer accuracy on XTREME-R when using English-only vs. multilingual intermediate training, and how does this compare to baseline models in terms of throughput and efficiency?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.0/10.