TabDSR: Decompose, Sanitize, and Reason for Complex Numerical Reasoning in Tabular Data

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Jiang, Changjiang, Yu, Fengchang, Chen, Haihua, Lu, Wei, Zeng, Jin
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911248923230208
author Jiang, Changjiang
Yu, Fengchang
Chen, Haihua
Lu, Wei
Zeng, Jin
author_facet Jiang, Changjiang
Yu, Fengchang
Chen, Haihua
Lu, Wei
Zeng, Jin
contents Complex reasoning over tabular data is crucial in real-world data analysis, yet large language models (LLMs) often underperform due to complex queries, noisy data, and limited numerical capabilities. To address these issues, we propose TabDSR, a framework consisting of: (1) a query decomposer that breaks down complex questions, (2) a table sanitizer that cleans and filters noisy tables, and (3) a program-of-thoughts (PoT)-based reasoner that generates executable code to derive the final answer from the sanitized table. To ensure unbiased evaluation and mitigate data leakage, we introduce a new dataset, CalTab151, specifically designed for complex numerical reasoning over tables. Experimental results demonstrate that TabDSR consistently outperforms existing methods, achieving state-of-the-art (SOTA) performance with 8.79%, 6.08%, and 19.87% accuracy improvement on TAT-QA, TableBench, and TabDSR, respectively. Moreover, our framework integrates seamlessly with mainstream LLMs, providing a robust solution for complex tabular numerical reasoning. These findings highlight the effectiveness of our framework in enhancing LLM performance for complex tabular numerical reasoning. Data and code are available upon request.
format Preprint
id arxiv_https___arxiv_org_abs_2511_02219
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle TabDSR: Decompose, Sanitize, and Reason for Complex Numerical Reasoning in Tabular Data
Jiang, Changjiang
Yu, Fengchang
Chen, Haihua
Lu, Wei
Zeng, Jin
Artificial Intelligence
Complex reasoning over tabular data is crucial in real-world data analysis, yet large language models (LLMs) often underperform due to complex queries, noisy data, and limited numerical capabilities. To address these issues, we propose TabDSR, a framework consisting of: (1) a query decomposer that breaks down complex questions, (2) a table sanitizer that cleans and filters noisy tables, and (3) a program-of-thoughts (PoT)-based reasoner that generates executable code to derive the final answer from the sanitized table. To ensure unbiased evaluation and mitigate data leakage, we introduce a new dataset, CalTab151, specifically designed for complex numerical reasoning over tables. Experimental results demonstrate that TabDSR consistently outperforms existing methods, achieving state-of-the-art (SOTA) performance with 8.79%, 6.08%, and 19.87% accuracy improvement on TAT-QA, TableBench, and TabDSR, respectively. Moreover, our framework integrates seamlessly with mainstream LLMs, providing a robust solution for complex tabular numerical reasoning. These findings highlight the effectiveness of our framework in enhancing LLM performance for complex tabular numerical reasoning. Data and code are available upon request.
title TabDSR: Decompose, Sanitize, and Reason for Complex Numerical Reasoning in Tabular Data
topic Artificial Intelligence
url https://arxiv.org/abs/2511.02219