GRADE: Benchmarking Discipline-Informed Reasoning in Image Editing

TL;DR

GRADE introduces a comprehensive benchmark for evaluating discipline-informed reasoning in image editing, revealing current models' limitations in knowledge-intensive, domain-specific tasks across multiple academic fields.

Contribution

This work presents GRADE, the first benchmark specifically designed to assess discipline-informed reasoning in image editing, with a multi-dimensional evaluation protocol and extensive experimental analysis.

Findings

Current models show significant performance gaps in discipline-informed image editing.

The benchmark reveals limitations in models' implicit, knowledge-intensive reasoning abilities.

Analysis exposes specific shortcomings and constraints in existing models' disciplinary understanding.

Abstract

Unified multimodal models target joint understanding, reasoning, and generation, but current image editing benchmarks are largely confined to natural images and shallow commonsense reasoning, offering limited assessment of this capability under structured, domain-specific constraints. In this work, we introduce GRADE, the first benchmark to assess discipline-informed knowledge and reasoning in image editing. GRADE comprises 520 carefully curated samples across 10 academic domains, spanning from natural science to social science. To support rigorous evaluation, we propose a multi-dimensional evaluation protocol that jointly assesses Discipline Reasoning, Visual Consistency, and Logical Readability. Extensive experiments on 20 state-of-the-art open-source and closed-source models reveal substantial limitations in current models under implicit, knowledge-intensive editing settings, leading to large performance gaps. Beyond quantitative scores, we conduct rigorous analyses and ablations to expose model shortcomings and identify the constraints within disciplinary editing. Together, GRADE pinpoints key directions for the future development of unified multimodal models, advancing the research on discipline-informed image editing and reasoning. Our benchmark and evaluation code are publicly released.