Hidden Meanings in Plain Sight: RebusBench for Evaluating Cognitive Visual Reasoning

TL;DR

RebusBench is a new benchmark designed to evaluate the ability of vision-language models to perform complex, multi-step cognitive reasoning required to solve rebus puzzles, highlighting current models' deficiencies in this area.

Contribution

The paper introduces RebusBench, a benchmark of 1,164 puzzles, and evaluates state-of-the-art models, revealing significant gaps in their cognitive visual reasoning capabilities.

Findings

Models perform below 10% exact match on RebusBench.

Scaling models and in-context learning do not significantly improve performance.

Current models lack the cognitive reasoning ability to connect perception and knowledge.

Abstract

Large Vision-Language Models (LVLMs) have achieved remarkable proficiency in explicit visual recognition, effectively describing what is directly visible in an image. However, a critical cognitive gap emerges when the visual input serves only as a clue rather than the answer. We identify that current models struggle with the complex, multi-step reasoning required to solve problems where information is not explicitly depicted. Successfully solving a rebus puzzle requires a distinct cognitive workflow: the model must extract visual and textual attributes, retrieve linguistic prior knowledge (such as idioms), and perform abstract mapping to synthesize these elements into a meaning that exists outside the pixel space. To evaluate this neurosymbolic capability, we introduce RebusBench, a benchmark of 1,164 puzzles designed to test this specific integration of perception and knowledge. Our evaluation of state-of-the-art models (including Qwen, InternVL, and LLaVA) shows a severe deficiency: performance saturates below 10% Exact Match and 20% semantic accuracy, with no significant improvement observed from model scaling or In-Context Learning (ICL). These findings suggest that while models possess the necessary visual and linguistic components, they lack the cognitive reasoning glue to connect them. Project page available at https://amirkasaei.com/rebusbench/.