VERIFY: A Benchmark of Visual Explanation and Reasoning for Investigating Multimodal Reasoning Fidelity
Jing Bi, Junjia Guo, Susan Liang, Guangyu Sun, Luchuan Song, Yunlong Tang, Jinxi He, Jiarui Wu, Ali Vosoughi, Chen Chen, Chenliang Xu
2025-03-20
Summary
This paper introduces a new test to see how well AI models can understand and reason about images, even without a lot of extra information.
What's the problem?
Current tests for AI models mostly check if they can recognize things in images, but not if they truly understand what they're seeing.
What's the solution?
The researchers created VERIFY, a test that forces AI models to reason based on images alone, with minimal text to help them.
Why it matters?
This work is important because it helps us understand if AI models are actually 'thinking' about images or just recognizing patterns.
Abstract
Visual reasoning is central to human cognition, enabling individuals to interpret and abstractly understand their environment. Although recent Multimodal Large Language Models (MLLMs) have demonstrated impressive performance across language and vision-language tasks, existing benchmarks primarily measure recognition-based skills and inadequately assess true visual reasoning capabilities. To bridge this critical gap, we introduce VERIFY, a benchmark explicitly designed to isolate and rigorously evaluate the visual reasoning capabilities of state-of-the-art MLLMs. VERIFY compels models to reason primarily from visual information, providing minimal textual context to reduce reliance on domain-specific knowledge and linguistic biases. Each problem is accompanied by a human-annotated reasoning path, making it the first to provide in-depth evaluation of model decision-making processes. Additionally, we propose novel metrics that assess visual reasoning fidelity beyond mere accuracy, highlighting critical imbalances in current model reasoning patterns. Our comprehensive benchmarking of leading MLLMs uncovers significant limitations, underscoring the need for a balanced and holistic approach to both perception and reasoning. For more teaser and testing, visit our project page (https://verify-eqh.pages.dev/).