UAQFact: Evaluating Factual Knowledge Utilization of LLMs on Unanswerable Questions

2025-05-29

AI Summary

Existing Unanswerable Question (UAQ) evaluation datasets lack factual knowledge grounding, hindering rigorous assessment of large language models’ (LLMs) ability to leverage factual knowledge in UAQ scenarios. Method: We introduce UAQFact—the first bilingual UAQ benchmark augmented with auxiliary factual knowledge, automatically constructed from knowledge graphs and covering both implicit and explicit reasoning paths. We propose a dual-task evaluation paradigm to separately measure models’ capability to retrieve internal parametric knowledge and integrate externally injected factual knowledge. Results: Experiments reveal substantial performance degradation across mainstream LLMs (e.g., Llama, Qwen, GLM) on UAQFact; external knowledge injection yields only marginal gains; and prevalent phenomena—including factual neglect and spurious associations—indicate fundamental deficiencies in LLMs’ factual knowledge utilization mechanisms.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Machine Learning: Large Multimodal Models (LMMs)
• Knowledge Representation and Reasoning: Knowledge Acquisition

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Handling unanswerable questions (UAQ) is crucial for LLMs, as it helps prevent misleading responses in complex situations. While previous studies have built several datasets to assess LLMs' performance on UAQ, these datasets lack factual knowledge support, which limits the evaluation of LLMs' ability to utilize their factual knowledge when handling UAQ. To address the limitation, we introduce a new unanswerable question dataset UAQFact, a bilingual dataset with auxiliary factual knowledge created from a Knowledge Graph. Based on UAQFact, we further define two new tasks to measure LLMs' ability to utilize internal and external factual knowledge, respectively. Our experimental results across multiple LLM series show that UAQFact presents significant challenges, as LLMs do not consistently perform well even when they have factual knowledge stored. Additionally, we find that incorporating external knowledge may enhance performance, but LLMs still cannot make full use of the knowledge which may result in incorrect responses.

Problem

• Evaluating LLMs' ability to handle unanswerable questions with factual knowledge
• Addressing lack of factual knowledge support in existing UAQ datasets
• Assessing LLMs' utilization of internal and external knowledge for UAQ

Innovation

• Introduces UAQFact dataset with Knowledge Graph
• Defines tasks for internal and external knowledge
• Evaluates LLMs' factual knowledge utilization challenges