Does QA-based intermediate training help fine-tuning language models for text classification?

Re-assign community