Factuality or Fiction? Benchmarking Modern LLMs on Ambiguous QA with Citations
This work addresses the problem of unreliable source citations and ambiguity handling in LLMs for QA systems, providing incremental benchmarking insights for developers.
The study evaluated the factual accuracy and citation performance of modern LLMs like GPT-4o-mini and Claude-3.5 on ambiguous QA tasks, finding that while models often predict at least one correct answer, they struggle with multiple valid answers and have citation accuracy of 0, but conflict-aware prompting significantly improves these aspects.
Benchmarking modern large language models (LLMs) on complex and realistic tasks is critical to advancing their development. In this work, we evaluate the factual accuracy and citation performance of state-of-the-art LLMs on the task of Question Answering (QA) in ambiguous settings with source citations. Using three recently published datasets-DisentQA-DupliCite, DisentQA-ParaCite, and AmbigQA-Cite-featuring a range of real-world ambiguities, we analyze the performance of two leading LLMs, GPT-4o-mini and Claude-3.5. Our results show that larger, recent models consistently predict at least one correct answer in ambiguous contexts but fail to handle cases with multiple valid answers. Additionally, all models perform equally poorly in citation generation, with citation accuracy consistently at 0. However, introducing conflict-aware prompting leads to large improvements, enabling models to better address multiple valid answers and improve citation accuracy, while maintaining their ability to predict correct answers. These findings highlight the challenges and opportunities in developing LLMs that can handle ambiguity and provide reliable source citations. Our benchmarking study provides critical insights and sets a foundation for future improvements in trustworthy and interpretable QA systems.