Reefknot: A Comprehensive Benchmark for Relation Hallucination Evaluation, Analysis and Mitigation in Multimodal Large Language Models

TL;DR

This work provides a systematic definition of relation hallucinations, integrating perceptive and cognitive perspectives, and construct a relation-based corpus using the Visual Genome scene graph dataset, and proposes a novel confidence-based mitigation strategy, which reduces the hallucination rate across three datasets, including Reefknot.

Abstract

Hallucination issues continue to affect multimodal large language models (MLLMs), with existing research mainly addressing object-level or attribute-level hallucinations, neglecting the more complex relation hallucinations that require advanced reasoning. Current benchmarks for relation hallucinations lack detailed evaluation and effective mitigation, and their datasets often suffer from biases due to systematic annotation processes. To address these challenges, we introduce Reefknot, a comprehensive benchmark targeting relation hallucinations, comprising over 20,000 real-world samples. We provide a systematic definition of relation hallucinations, integrating perceptive and cognitive perspectives, and construct a relation-based corpus using the Visual Genome scene graph dataset. Our comparative evaluation reveals significant limitations in current MLLMs' ability to handle relation hallucinations. Additionally, we propose a novel confidence-based mitigation strategy, which reduces the hallucination rate by an average of 9.75% across three datasets, including Reefknot. Our work offers valuable insights for achieving trustworthy multimodal intelligence.

Figures (18)

• Figure 1: Comparison between the focus of Reefknot — relation hallucination with two categories (i.e., perceptive & cognitive) vs. object & attribute hallucinations.
• Figure 2: The hallucination rates on POPE, an object hallucination benchmark, and our Reefknot with a focus on relation hallucination (w/ same configuration).
• Figure 3: The data construction pipeline of our proposed Reefknot benchmark.
• Figure 4: MLLMs with top five best performance evaluated on Reefknot benchmark. We report the sum of the respective metric across three tasks for reference.
• Figure 5: Confusion matrices of MiniCPM-7B's performance on Reefknot benchmark (Left: Y/N setting; Right: MCQ setting).