Haobai Sun, Ying Gu

Ying Gu

December 30, 2025

Large Language Models, Text generation, Automated proofreading, Retrieval-augmented generation, Domain adaptation, Ethical AI, Responsible automation.

The rapid expansion of digital communication has transformed how knowledge, policy, and industry operate, creating both unprecedented opportunity and substantial risk. As written text becomes the backbone of digital interaction, errors in clarity, factual accuracy, or coherence carry outsized consequences—from misinformation and misdiagnosis to legal disputes and financial loss. Traditional natural language processing (NLP) methods, such as rule-based or statistical systems, lack contextual understanding, while even advanced Large Language Models (LLMs) like GPT-4, Claude, and LLaMA remain vulnerable to factual hallucination, domain drift, and bias. This paper introduces the Unified Generation–Proofreading Framework (UGPF), a system that merges text generation, automated proofreading, and iterative refinement into a cohesive feedback loop. Drawing on retrieval-augmented generation (RAG), parameter-efficient domain tuning (LoRA), and ethical alignment via Constitutional AI, UGPF enables language models to generate text that is not only fluent but verifiably accurate and ethically sound, achieving 20–30% higher factual consistency scores compared to baselines. The framework unites theoretical principles with applied strategies, establishing a dual objective paradigm to systematically balance fluency with factual integrity and ethical compliance, thereby setting a new standard for trust and rigor in AI-generated content.

© 2025, the Authors. Published by ISCCAC

This is an open access article distributed under the CC BY-NC license