Truth or Mirage? Towards End-to-End Factuality Evaluation with LLM-Oasis

2024-11-29

arXiv.org

AI Summary

Large language models (LLMs) frequently generate factually inaccurate content (“hallucinations”), yet existing factual evaluation benchmarks suffer from task specificity, limited scale, and simplistic verification protocols. Method: We introduce LLM-Oasis—the first large-scale, task-agnostic, end-to-end factual evaluation benchmark—constructed by extracting real-world statements from Wikipedia, generating high-quality true/false textual pairs via controllable semantic perturbation, and establishing a gold-standard test set through multi-round expert annotation. Contribution/Results: LLM-Oasis enables cross-task generalization evaluation for factual consistency, offering high reliability and strong challenge (e.g., GPT-4o achieves only 60% accuracy). Our methodology integrates statement extraction, controlled falsification, multi-level human annotation, and LLM-driven factuality discrimination modeling, providing a scalable, robust benchmark for training and evaluating factual assessment models.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Evaluation and Analysis

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

After the introduction of Large Language Models (LLMs), there have been substantial improvements in the performance of Natural Language Generation (NLG) tasks, including Text Summarization and Machine Translation. However, LLMs still produce outputs containing hallucinations, that is, content not grounded in factual information. Therefore, developing methods to assess the factuality of LLMs has become urgent. Indeed, resources for factuality evaluation have recently emerged. Although challenging, these resources face one or more of the following limitations: (i) they are tailored to a specific task or domain; (ii) they are limited in size, thereby preventing the training of new factuality evaluators; (iii) they are designed for simpler verification tasks, such as claim verification. To address these issues, we introduce LLM-Oasis, to the best of our knowledge the largest resource for training end-to-end factuality evaluators. LLM-Oasis is constructed by extracting claims from Wikipedia, falsifying a subset of these claims, and generating pairs of factual and unfactual texts. We then rely on human annotators to both validate the quality of our dataset and to create a gold standard test set for benchmarking factuality evaluation systems. Our experiments demonstrate that LLM-Oasis presents a significant challenge for state-of-the-art LLMs, with GPT-4o achieving up to 60% accuracy in our proposed end-to-end factuality evaluation task, highlighting its potential to drive future research in the field.

Problem

• Evaluating factuality of LLM outputs to reduce hallucinations
• Overcoming limitations in existing factuality evaluation resources
• Creating a large dataset for training end-to-end factuality evaluators

Innovation

• Largest resource for end-to-end factuality evaluators
• Claims extracted and falsified from Wikipedia
• Human-annotated gold standard for benchmarking