MRT at IberLEF-2025 PRESTA Task: Maximizing Recovery from Tables with Multiple Steps

2025-07-17

AI Summary

This work addresses the Spanish-language table question answering (TQA) task by proposing a multi-stage natural language-to-code reasoning framework. Methodologically, it extends the Multi-Step Reasoning and Translation (MRT) framework with a stepwise inference pipeline: table structure understanding, key column identification, natural language instruction generation, Python code synthesis, and safe execution—augmented by fine-grained prompt optimization and robust exception handling. The core contribution lies in decomposing complex table semantic parsing into verifiable, intermediate reasoning steps, thereby significantly improving code generation accuracy and interpretability of open-source large language models (LLMs) for low-resource languages. Evaluated on the IberLEF-2025 PRESTA benchmark, the framework achieves 85% exact-match accuracy, outperforming existing end-to-end approaches.

Technology Category

• Natural Language Processing: Question Answering
• Knowledge Representation and Reasoning: Knowledge Representation Languages
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

This paper presents our approach for the IberLEF 2025 Task PRESTA: Preguntas y Respuestas sobre Tablas en Español (Questions and Answers about Tables in Spanish). Our solution obtains answers to the questions by implementing Python code generation with LLMs that is used to filter and process the table. This solution evolves from the MRT implementation for the Semeval 2025 related task. The process consists of multiple steps: analyzing and understanding the content of the table, selecting the useful columns, generating instructions in natural language, translating these instructions to code, running it, and handling potential errors or exceptions. These steps use open-source LLMs and fine-grained optimized prompts for each step. With this approach, we achieved an accuracy score of 85% in the task.

Problem

• Generate Python code from tables for Q&A
• Filter and process Spanish table data
• Optimize multi-step LLM table analysis

Innovation

• Python code generation with LLMs
• Fine-grained optimized prompts
• Multi-step table processing