Evaluating LLMs on Document-Based QA: Exact Answer Selection and Numerical Extraction using Cogtale dataset

TL;DR

This paper evaluates large language models on document-based QA tasks that require exact answer selection and numerical extraction using the CogTale dataset. Employing a retrieval-then-answer pipeline, zero-shot prompts, and GPT-4 and GPT-3.5-turbo, the authors benchmark performance across Yes-No, single-choice, single-choice-number, multiple-choice, and number-extraction questions. GPT-4 consistently outperforms GPT-3.5-turbo (41.84% vs 31.45% overall) but both models struggle with multiple-choice and numerical extraction, highlighting limitations for precise information extraction in real-world document analysis. The work underscores the importance of effective retrievers and prompting strategies and provides a framework for ongoing evaluation of LLMs in information retrieval and evidence synthesis contexts.

Abstract

Document-based Question-Answering (QA) tasks are crucial for precise information retrieval. While some existing work focus on evaluating large language models performance on retrieving and answering questions from documents, assessing the LLMs performance on QA types that require exact answer selection from predefined options and numerical extraction is yet to be fully assessed. In this paper, we specifically focus on this underexplored context and conduct empirical analysis of LLMs (GPT-4 and GPT-3.5) on question types, including single-choice, yes-no, multiple-choice, and number extraction questions from documents in zero-shot setting. We use the CogTale dataset for evaluation, which provide human expert-tagged responses, offering a robust benchmark for precision and factual grounding. We found that LLMs, particularly GPT-4, can precisely answer many single-choice and yes-no questions given relevant context, demonstrating their efficacy in information retrieval tasks. However, their performance diminishes when confronted with multiple-choice and number extraction formats, lowering the overall performance of the model on this task, indicating that these models may not yet be sufficiently reliable for the task. This limits the applications of LLMs on applications demanding precise information extraction from documents, such as meta-analysis tasks. These findings hinge on the assumption that the retrievers furnish pertinent context necessary for accurate responses, emphasizing the need for further research. Our work offers a framework for ongoing dataset evaluation, ensuring that LLM applications for information retrieval and document analysis continue to meet evolving standards.

Figures (1)

• Figure 1: Question-Answering (QA) Framework using LLM for a document-based QA task.