Cross-lingual Transfer Performance and Inference Efficiency Trade-offs in Intermediate-task Training

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: What is the trade-off between cross-lingual transfer performance and inference efficiency when using intermediate-task training on models with varying parameter sizes (e.g., 10B vs. 100B) in XTREME-L?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.7/10.