Large Language Models (LLMs) have achieved remarkable results in natural language generation, yet challenges remain in data-to-text (D2T) tasks, particularly in controlling output, ensuring transparency, and maintaining factual consistency with the input. We introduce the first LLM-based multi-agent framework for D2T generation, coordinating specialized agents to produce high-quality, interpretable outputs. Our system combines the reasoning and acting abilities of ReAct agents, the self-correction of Reflexion agents, and the quality assurance of Guardrail agents, all directed by an Orchestrator agent that assigns tasks to three specialists—content ordering, text structuring, and surface realization—and iteratively refines outputs based on Guardrail feedback. This closed-loop design enables precise control and dynamic optimization, yielding text that is coherent, accurate, and grounded in the input data. On a relatively simple dataset like WebNLG, our framework performs competitively with end-to-end systems, highlighting its promise for more complex D2T scenarios.

Previous studies have highlighted the advantages of pipeline neural architectures over end-to-end models, particularly in reducing text hallucination. In this study, we extend prior research by integrating pretrained language models (PLMs) into a pipeline framework, using both fine-tuning and prompting methods. Our findings show that fine-tuned PLMs consistently generate high quality text, especially within end-to-end architectures and at intermediate stages of the pipeline across various domains. These models also outperform prompt-based ones on automatic evaluation metrics but lag in human evaluations. Compared to the standard five-stage pipeline architecture, a streamlined three-stage pipeline, which only include ordering, structuring, and surface realization, achieves superior performance in fluency and semantic adequacy according to the human evaluation.

Neural end-to-end surface realizers output more fluent texts than classical architectures. However, they tend to suffer from adequacy problems, in particular hallucinations in numerical referring expression generation. This poses a problem to language generation in sensitive domains, as is the case of robot journalism covering COVID-19 and Amazon deforestation. We propose an approach whereby numerical referring expressions are converted from digits to plain word form descriptions prior to being fed to state-of-the-art Large Language Models. We conduct automatic and human evaluations to report the best strategy to numerical superficial realization. Code and data are publicly available.