Fine-tuning and prompting are two widely used approaches for adapting large language models to downstream tasks. However, their effectiveness on text simplification remains underexplored. We present SIMPLE, a comprehensive evaluation of large language model-based sentence simplification. We evaluate 12 open-weight large language models across diverse domains, including Wikipedia, medical, scientific, and news texts. In addition to standard evaluation metrics, we conduct a linguistic analysis of simplification outputs at the lexical, syntactic, and discourse levels. Our findings show that: 1) under practical deployment settings, QLoRA fine-tuning of base models performs better on Wikipedia-based simplification, whereas optimized prompting of instruction-tuned models performs better in non-Wikipedia domains; 2) QLoRA-fine-tuned models exhibit a strong alignment with human-written references across target readability levels, while prompted models show limited controllability over output difficulty; and 3) the primary discrepancies between large language model-generated simplifications and human references arise from syntactic restructuring and discourse coherence.

Lexical-level sentence simplification is essential for improving text accessibility, yet traditional methods often struggle to dynamically identify complex terms and generate contextually appropriate substitutions, resulting in limited generalization. While prompt-based approaches with large language models (LLMs) have shown strong performance and adaptability, they often lack interpretability and are prone to hallucinating. This study proposes a fine-tuning approach for mid-sized LLMs to emulate the lexical simplification pipeline. We transform complex word identification datasets into an instruction–response format to support instruction tuning. Experimental results show that our method substantially enhances complex word identification accuracy with reduced hallucinations while achieving competitive performance on lexical simplification benchmarks. Furthermore, we find that integrating fine-tuning with prompt engineering reduces dependency on manual prompt optimization, leading to a more efficient simplification framework.