PhyEditBench: A Real-World Multi-Stage Benchmark for Physics-Aware Image Editing
Quick Answer
PhyEditBench introduces a benchmark for evaluating physics-aware image editing models, featuring 238 real-world instances and 35 synthetic cases.
Quick Take
PhyEditBench introduces a benchmark for evaluating physics-aware image editing models, featuring 238 real-world instances and 35 synthetic cases. The study reveals significant limitations in current state-of-the-art methods, while the proposed PhyWorld baseline demonstrates superior performance through innovative reasoning mechanisms.
Key Points
- PhyEditBench consists of 238 real-world and 35 synthetic instances for evaluation.
- Current state-of-the-art editing methods show significant limitations in physics-based reasoning.
- PhyWorld, a training-free baseline, utilizes test-time scaling for improved performance.
- The benchmark aims to enhance the physical understanding in image editing models.
- Project details are available at https://github.com/Previsior/PhyEditBench.
Paper Resources
📖 Reader Mode
~2 min readAbstract:While instruction-based image editing, enabled by multi-modal generative models, has advanced significantly, existing benchmarks lack a comprehensive evaluation of physics-based reasoning, a critical capability for handling real-world scenarios. To address this, we introduce PhyEditBench, a benchmark designed to assess the physical understanding of editing models. Guided by a hierarchical taxonomy, we establish 4 primary classes and 12 subclasses. It comprises 238 high-quality, high-resolution, real-world instances meticulously extracted from videos to capture authentic physical dynamics, alongside 35 synthetic Anti-Physics instances. Our empirical analysis of current SOTA editing methods exposes substantial limitations in their physics-based reasoning. We further propose a training-free baseline named PhyWorld that uses test-time scaling and a latent reduction strategy. PhyWorld outperforms comparable models and suggests that the video generation process can effectively serve as a reasoning mechanism for image editing. The project page is available at this https URL.
| Comments: | 19 pages, 6 figures, 2 tables. Accepted to ECCV 2026 |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2606.26551 [cs.CV] |
| (or arXiv:2606.26551v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2606.26551 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Shengbin Guo [view email]
[v1]
Thu, 25 Jun 2026 02:57:34 UTC (6,248 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →LLM-Guided ANN Index Optimization for Human-Object Interaction Retrieval
A phase-aware LLM agent optimizes human-object interaction retrieval, outperforming Optuna TPE by 33.3% and VDTuner by 34.2% on the HICO-DET benchmark. This method enhances throughput by 15.3x over UniIR and demonstrates strong transferability across vector database management systems.