LyS at SemEval 2025 Task 8: Zero-Shot Code Generation for Tabular QA

TL;DR

This paper presents a zero-shot code generation pipeline using large language models for tabular question answering, demonstrating its viability despite no task-specific fine-tuning.

Contribution

It introduces a modular zero-shot approach with iterative refinement for tabular QA, showing competitive results without fine-tuning.

Findings

Achieved rank 33 out of 53 in SemEval 2025 Task 8.

Demonstrated the effectiveness of zero-shot code generation for Tabular QA.

Implemented an iterative refinement process to improve robustness.

Abstract

This paper describes our participation in SemEval 2025 Task 8, focused on Tabular Question Answering. We developed a zero-shot pipeline that leverages an Large Language Model to generate functional code capable of extracting the relevant information from tabular data based on an input question. Our approach consists of a modular pipeline where the main code generator module is supported by additional components that identify the most relevant columns and analyze their data types to improve extraction accuracy. In the event that the generated code fails, an iterative refinement process is triggered, incorporating the error feedback into a new generation prompt to enhance robustness. Our results show that zero-shot code generation is a valid approach for Tabular QA, achieving rank 33 of 53 in the test phase despite the lack of task-specific fine-tuning.