OUNLP at TSAR 2025 Shared Task: Multi-Round Text Simplifier via Code Generation

TL;DR

This paper presents a multi-round text simplification system using GPT-4o, which leverages CEFR level gaps and rule-based methods to improve readability, achieving competitive results in the TSAR-2025 Shared Task.

Contribution

It introduces two novel multi-round simplification methods based on prompt-based generation and demonstrates their effectiveness for controlled text simplification.

Findings

Performance correlates with CEFR level gaps.

Multi-round methods outperform single-pass approaches.

Using LLM-generated candidates as starting points boosts results.

Abstract

This paper describes the OUNLP system submitted to the TSAR-2025 Shared Task (Alva-Manchego et al., 2025), designed for readability-controlled text simplification using LLM-prompting-based generation. Based on the analysis of prompt-based text simplification methods, we discovered an interesting finding that text simplification performance is highly related to the gap between the source CEFR (Arase et al., 2022) level and the target CEFR level. Inspired by this finding, we propose two multi-round simplification methods and generate them via GPT-4o: rule-based simplification (MRS-Rule) and jointly rule-based LLM simplification (MRS-Joint). Our submitted systems ranked 7 out of 20 teams. Later improvements with MRS-Joint show that taking the LLM simplified candidates as the starting point could further boost the multi-round simplification performance.