JarvisEvo: Towards a Self-Evolving Photo Editing Agent with Synergistic Editor-Evaluator Optimization

TL;DR

JarvisEvo introduces a self-evolving photo editing agent that combines multimodal reasoning and a synergistic policy framework to improve editing accuracy and reduce reward hacking, outperforming existing models on key metrics.

Contribution

It presents a novel unified agent with interleaved multimodal reasoning and self-improvement capabilities, advancing photo editing AI beyond prior instruction-following models.

Findings

Outperforms Nano-Banana by 18.95% on editing metrics

Achieves 44.96% improvement in pixel-level content fidelity

Demonstrates effective self-improvement without external rewards

Abstract

Agent-based editing models have substantially advanced interactive experiences, processing quality, and creative flexibility. However, two critical challenges persist: (1) instruction hallucination, text-only chain-of-thought (CoT) reasoning cannot fully prevent factual errors due to inherent information bottlenecks; (2) reward hacking, dynamic policy optimization against static reward models allows agents to exploit flaws in reward functions. To address these issues, we propose JarvisEvo, a unified image editing agent that emulates an expert human designer by iteratively editing, selecting appropriate tools, evaluating results, and reflecting on its own decisions to refine outcomes. JarvisEvo offers three key advantages: (1) an interleaved multimodal chain-of-thought (iMCoT) reasoning mechanism that enhances instruction following and editing quality; (2) a synergistic editor-evaluator policy optimization (SEPO) framework that enables self-improvement without external rewards, effectively mitigating reward hacking; and (3) support for both global and local fine-grained editing through seamless integration of Adobe Lightroom. On ArtEdit-Bench, JarvisEvo outperforms Nano-Banana by an average of 18.95% on preservative editing metrics, including a substantial 44.96% improvement in pixel-level content fidelity. Project page: https://jarvisevo.vercel.app/