Exploring the Robustness of Language Models for Tabular Question Answering via Attention Analysis

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Bhandari, Kushal Raj, Xing, Sixue, Dan, Soham, Gao, Jianxi
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915463865303040
author Bhandari, Kushal Raj
Xing, Sixue
Dan, Soham
Gao, Jianxi
author_facet Bhandari, Kushal Raj
Xing, Sixue
Dan, Soham
Gao, Jianxi
contents Large Language Models (LLMs), already shown to ace various unstructured text comprehension tasks, have also remarkably been shown to tackle table (structured) comprehension tasks without specific training. Building on earlier studies of LLMs for tabular tasks, we probe how in-context learning (ICL), model scale, instruction tuning, and domain bias affect Tabular QA (TQA) robustness by testing LLMs, under diverse augmentations and perturbations, on diverse domains: Wikipedia-based $\textbf{WTQ}$, financial $\textbf{TAT-QA}$, and scientific $\textbf{SCITAB}$. Although instruction tuning and larger, newer LLMs deliver stronger, more robust TQA performance, data contamination and reliability issues, especially on $\textbf{WTQ}$, remain unresolved. Through an in-depth attention analysis, we reveal a strong correlation between perturbation-induced shifts in attention dispersion and the drops in performance, with sensitivity peaking in the model's middle layers. We highlight the need for improved interpretable methodologies to develop more reliable LLMs for table comprehension. Through an in-depth attention analysis, we reveal a strong correlation between perturbation-induced shifts in attention dispersion and performance drops, with sensitivity peaking in the model's middle layers. Based on these findings, we argue for the development of structure-aware self-attention mechanisms and domain-adaptive processing techniques to improve the transparency, generalization, and real-world reliability of LLMs on tabular data.
format Preprint
id arxiv_https___arxiv_org_abs_2406_12719
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Exploring the Robustness of Language Models for Tabular Question Answering via Attention Analysis
Bhandari, Kushal Raj
Xing, Sixue
Dan, Soham
Gao, Jianxi
Computation and Language
Artificial Intelligence
Large Language Models (LLMs), already shown to ace various unstructured text comprehension tasks, have also remarkably been shown to tackle table (structured) comprehension tasks without specific training. Building on earlier studies of LLMs for tabular tasks, we probe how in-context learning (ICL), model scale, instruction tuning, and domain bias affect Tabular QA (TQA) robustness by testing LLMs, under diverse augmentations and perturbations, on diverse domains: Wikipedia-based $\textbf{WTQ}$, financial $\textbf{TAT-QA}$, and scientific $\textbf{SCITAB}$. Although instruction tuning and larger, newer LLMs deliver stronger, more robust TQA performance, data contamination and reliability issues, especially on $\textbf{WTQ}$, remain unresolved. Through an in-depth attention analysis, we reveal a strong correlation between perturbation-induced shifts in attention dispersion and the drops in performance, with sensitivity peaking in the model's middle layers. We highlight the need for improved interpretable methodologies to develop more reliable LLMs for table comprehension. Through an in-depth attention analysis, we reveal a strong correlation between perturbation-induced shifts in attention dispersion and performance drops, with sensitivity peaking in the model's middle layers. Based on these findings, we argue for the development of structure-aware self-attention mechanisms and domain-adaptive processing techniques to improve the transparency, generalization, and real-world reliability of LLMs on tabular data.
title Exploring the Robustness of Language Models for Tabular Question Answering via Attention Analysis
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2406.12719