V-REX: Benchmarking Exploratory Visual Reasoning via Chain-of-Questions

TL;DR

V-REX introduces a comprehensive benchmark and evaluation protocol for multi-step visual reasoning, emphasizing planning and following capabilities in vision-language models through a chain-of-questions approach.

Contribution

The paper presents V-REX, a novel benchmark and evaluation framework for multi-step exploratory visual reasoning, enabling detailed assessment of models' planning and following skills.

Findings

Significant performance gaps in current models' multi-step reasoning abilities.

Clear scaling trends observed across different models.

Notable differences between planning and following capabilities.

Abstract

While many vision-language models (VLMs) are developed to answer well-defined, straightforward questions with highly specified targets, as in most benchmarks, they often struggle in practice with complex open-ended tasks, which usually require multiple rounds of exploration and reasoning in the visual space. Such visual thinking paths not only provide step-by-step exploration and verification as an AI detective but also produce better interpretations of the final answers. However, these paths are challenging to evaluate due to the large exploration space of intermediate steps. To bridge the gap, we develop an evaluation suite, ``Visual Reasoning with multi-step EXploration (V-REX)'', which is composed of a benchmark of challenging visual reasoning tasks requiring native multi-step exploration and an evaluation protocol. V-REX covers rich application scenarios across diverse domains. V-REX casts the multi-step exploratory reasoning into a Chain-of-Questions (CoQ) and disentangles VLMs' capability to (1) Planning: breaking down an open-ended task by selecting a chain of exploratory questions; and (2) Following: answering curated CoQ sequentially to collect information for deriving the final answer. By curating finite options of questions and answers per step, V-REX achieves a reliable quantitative and fine-grained analysis of the intermediate steps. By assessing SOTA proprietary and open-sourced VLMs, we reveal consistent scaling trends, significant differences between planning and following abilities, and substantial room for improvement in multi-step exploratory reasoning.