ACL Findings2025
A Constrained Text Revision Agent via Iterative Planning and Searching
Hannan Cao, Hwee Tou Ng
Abstract
Existing text revision systems are capable of generating fluent and coherent text, but struggle with constrained text revision (CTR), which requires adherence to specific constraints. Furthermore, adapting these systems to diverse constraints is challenging. To bridge this gap, we introduce CRAFT, a Constrained Revision Agent For Text, focusing on CTR. CRAFT utilizes a planner, a reviser (i.e., a large language model), and adaptable tools to generate revisions tailored to different scenarios. Specifically, we propose an iterative self-training alignment method to construct the planner, which generates tool usage and text revision plans. Furthermore, we propose Tool-Guided Monte Carlo Tree Search (TG-MCTS), a novel CTR algorithm that extends MCTS with toolguided expansion and evaluation, enabling the search for optimal revision strategies across various scenarios. To evaluate CRAFT, we introduce CORD (COnstrained Revision Dataset), a dataset with multi-level constrained instructions for paragraph-level revision. Experimental results show that CRAFT outperforms baselines in both constraint adherence and revision quality. Furthermore, CRAFT exhibits robust performance across diverse use cases, including plain text and LaTeX revision. 1
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
Builds on22
- Chain-of-Thought Prompting Elicits Reasoning in Large Language ModelsJason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma et al.NeurIPS 2022 · 22,562 citations
- Toolformer: Language Models Can Teach Themselves to Use ToolsTimo Schick, Jane Dwivedi-Yu, Roberto Dessì, Roberta Raileanu et al.NeurIPS 2023 · 5,989 citations
- WizardLM: Empowering Large Pre-Trained Language Models to Follow Complex InstructionsCan Xu, Qingfeng Sun, Kai Zheng, Xiubo Geng et al.ICLR 2024 · 1,206 citations
- SimPO: Simple Preference Optimization with a Reference-Free RewardYu Meng, Mengzhou Xia, Danqi ChenNeurIPS 2024 · 1,203 citations
- BARTScore: Evaluating Generated Text as Text GenerationWeizhe Yuan, Graham Neubig, Pengfei LiuNeurIPS 2021 · 1,143 citations
Related papers
- ToC: Tree-of-Claims Search with Multi-Agent Language ModelsShuyang Yu, Jianan Liang, Hui HuAAAI 2026
- TOOLCAD: Exploring Tool-Using Large Language Models in Text-to-CAD Generation with Reinforcement LearningXing Wu Yifei Gong, Tukang Wenda LiuACL Findings 2026 · 1 citation
- SuperWriter: Reflection-Driven Long-Form Generation with Large Language ModelsYuhao Wu, Yushi Bai, Zhiqiang Hu, Juanzi Li et al.ACL Findings 2026 · 1 citation
- ToolTree: Efficient LLM Tool Planning via Dual-Feedback Monte Carlo Tree Search and Bidirectional PruningShuo Yang, Caren Han, Yihao Ding, Shuhe Wang et al.ICLR 2026 · 9 citations
- CARL: Constraint-Aware Reinforcement Learning for Planning with LLMsQiuyi Qi, Jinjian Zhang, Mutian Bao, Tian Liang et al.ACL Findings 2026
