KRIS-Bench: Benchmarking Next-Level Intelligent Image Editing Models

TL;DR

KRIS-Bench is a new benchmark for evaluating the reasoning capabilities of image editing models across factual, conceptual, and procedural knowledge, revealing significant gaps in current models' reasoning abilities.

Contribution

The paper introduces KRIS-Bench, a comprehensive, knowledge-based benchmark with 22 tasks and a novel evaluation protocol for assessing reasoning in image editing models.

Findings

Current models show significant reasoning gaps.

KRIS-Bench reveals limitations in knowledge-based editing.

Benchmark facilitates targeted improvements in model reasoning.

Abstract

Recent advances in multi-modal generative models have enabled significant progress in instruction-based image editing. However, while these models produce visually plausible outputs, their capacity for knowledge-based reasoning editing tasks remains under-explored. In this paper, we introduce KRIS-Bench (Knowledge-based Reasoning in Image-editing Systems Benchmark), a diagnostic benchmark designed to assess models through a cognitively informed lens. Drawing from educational theory, KRIS-Bench categorizes editing tasks across three foundational knowledge types: Factual, Conceptual, and Procedural. Based on this taxonomy, we design 22 representative tasks spanning 7 reasoning dimensions and release 1,267 high-quality annotated editing instances. To support fine-grained evaluation, we propose a comprehensive protocol that incorporates a novel Knowledge Plausibility metric, enhanced by knowledge hints and calibrated through human studies. Empirical results on 10 state-of-the-art models reveal significant gaps in reasoning performance, highlighting the need for knowledge-centric benchmarks to advance the development of intelligent image editing systems.