QASE Enhanced PLMs: Improved Control in Text Generation for MRC
This addresses control issues in text generation for MRC, which is a domain-specific problem, and appears incremental as it builds on existing PLMs.
The paper tackles the problem of out-of-control generation in generative models for machine reading comprehension (MRC) by introducing a Question-Attended Span Extraction (QASE) module during fine-tuning, resulting in performance matching SOTA extractive methods and outperforming leading LLMs like GPT-4 in MRC tasks.
To address the challenges of out-of-control generation in generative models for machine reading comprehension (MRC), we introduce the Question-Attended Span Extraction (QASE) module. Integrated during the fine-tuning of pre-trained generative language models (PLMs), QASE enables these PLMs to match SOTA extractive methods and outperform leading LLMs like GPT-4 in MRC tasks, without significant increases in computational costs.