LLaM-ScoreNet: An Advanced Architecture for Automated Text Quality Assessment Based on LLaMA Language Model

This paper presents a new approach for automating text quality evaluation using LLaM-ScoreNet, a framework built on the LLaMA pre-trained language model. The architecture integrates a Dynamic Attention Mechanism (DAM), a Dual-Branch Scoring System, and an objective evaluation module to improve fluency, normativity, and overall text quality assessment. By using deep semantic representations and focusing on key linguistic regions, LLaM-ScoreNet enhances evaluation accuracy and reliability. The dual-branch system evaluates fluency and normativity separately, and the final score is derived from a weighted fusion of both outputs. The objective evaluation module ensures correct answer assessments. Experimental results show that LLaM-ScoreNet outperforms existing models, demonstrating better performance on key metrics and confirming the importance of each component.