Intermediate-Task Dataset Scaling and Zero-Shot Cross-Lingual Transfer Performance

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: How does scaling the intermediate-task training dataset size affect the zero-shot cross-lingual transfer performance on XTREME-R tasks, specifically comparing accuracy gains when using intermediate datasets of 10K, 50K, and 100K English examples?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.8/10.