Multimodal vs. Text-Only Intermediate-Task Training for Zero-Shot Cross-Lingual Visual Language Model 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 intermediate-task training on multimodal datasets like VQAv2 compare to text-only intermediate tasks in enhancing zero-shot cross-lingual performance on visual language models, measured by VQA accuracy across languages?

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