How Well Do Models Follow Visual Instructions? VIBE: A Systematic Benchmark for Visual Instruction-Driven Image Editing

TL;DR

VIBE introduces a comprehensive benchmark for evaluating how well models follow complex visual instructions in image editing, revealing performance gaps and guiding future improvements.

Contribution

The paper presents VIBE, a new systematic benchmark with a hierarchical interaction framework and a robust evaluation method for visual instruction-driven image editing models.

Findings

Proprietary models outperform open-source counterparts.

Model performance declines with increasing task complexity.

VIBE enables detailed assessment of visual instruction-following capabilities.

Abstract

Recent generative models have achieved remarkable progress in image editing. However, existing systems and benchmarks remain largely text-guided. In contrast, human communication is inherently multimodal, where visual instructions such as sketches efficiently convey spatial and structural intent. To address this gap, we introduce VIBE, the Visual Instruction Benchmark for Image Editing with a three-level interaction hierarchy that captures deictic grounding, morphological manipulation, and causal reasoning. Across these levels, we curate high-quality and diverse test cases that reflect progressively increasing complexity in visual instruction following. We further propose a robust LMM-as-a-judge evaluation framework with task-specific metrics to enable scalable and fine-grained assessment. Through a comprehensive evaluation of 17 representative open-source and proprietary image editing models, we find that proprietary models exhibit early-stage visual instruction-following capabilities and consistently outperform open-source models. However, performance degrades markedly with increasing task difficulty even for the strongest systems, highlighting promising directions for future research.