FIHA: Autonomous Hallucination Evaluation in Vision-Language Models with Davidson Scene Graphs

2024-09-20

arXiv.org

AI Summary

Evaluating hallucinations in Large Vision-Language Models (LVLMs) is costly and lacks comprehensive coverage. Method: This paper proposes FIHA—a fine-grained, LLM-free, and annotation-free automatic evaluation framework. It introduces the first structured modeling of hallucination type dependencies using Davidson Scene Graphs (DSGs); designs a dual-path (image- and text-based) assessment mechanism; develops a three-dimensional hallucination detection model grounded in relations, attributes, and dependencies; and enhances cross-domain generalization by integrating MSCOCO and Foggy datasets. Contribution/Results: We release FIHA-v1—the first fine-grained, structured hallucination benchmark—and systematically uncover critical deficiencies of mainstream LVLMs in relational and attribute dependency reasoning. All code and data are open-sourced to advance low-cost, reproducible hallucination evaluation.

Technology Category

• Computer Vision: Large Vision Models
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

The rapid development of Large Vision-Language Models (LVLMs) often comes with widespread hallucination issues, making cost-effective and comprehensive assessments increasingly vital. Current approaches mainly rely on costly annotations and are not comprehensive -- in terms of evaluating all aspects such as relations, attributes, and dependencies between aspects. Therefore, we introduce the FIHA (autonomous Fine-graIned Hallucination evAluation evaluation in LVLMs), which could access hallucination LVLMs in the LLM-free and annotation-free way and model the dependency between different types of hallucinations. FIHA can generate Q&A pairs on any image dataset at minimal cost, enabling hallucination assessment from both image and caption. Based on this approach, we introduce a benchmark called FIHA-v1, which consists of diverse questions on various images from MSCOCO and Foggy. Furthermore, we use the Davidson Scene Graph (DSG) to organize the structure among Q&A pairs, in which we can increase the reliability of the evaluation. We evaluate representative models using FIHA-v1, highlighting their limitations and challenges. We released our code and data.

Problem

• Evaluating hallucination issues in LVLMs cost-effectively and comprehensively
• Assessing relations, attributes, and dependencies between aspects in LVLMs
• Providing an LLM-free and annotation-free hallucination evaluation method

Innovation

• Autonomous hallucination evaluation without LLM
• Generates Q&A pairs at minimal cost
• Uses Davidson Scene Graph for reliability