Adobe Research Unveils EditAgent for Precise AI Image Editing

Adobe Research and UC San Diego built EditAgent, an AI system designed to break image editing tasks into sub-steps. The system pairs a general-purpose editor with specialized tools for functions like sharpening, rotation, and zooming. A standard language model evaluates image variants generated in parallel using measured statistics and close-up crops. This evaluation model outperforms models specifically trained for judging image edits. In a test of 600 image edits, a general-purpose editor failed to produce a usable result 42.7 percent of the time. Specialized tools solved roughly two-thirds of these failures. Professionals preferred parallel variants generated from the original image over iterative refinement in 78.5 percent of comparisons.

For each variant, the EditAgent model receives a verification record containing a close-up, a change map, and measured statistics on brightness and color. With this record, all seven tested language models beat every fine-tuned model in picking usable edits. The Qwen3.8-27B model made a usable pick 81.3 percent of the time. EditAgent checks whether later edits accidentally undo earlier ones. The system improved the open-source editors Qwen-Image-Edit-2511 and Flux-2-Klein-9B across three benchmarks. On the EdiVal benchmark, the Qwen variant in its third editing round edged past the commercial models Seedream 4.0 and GPT-Image-1.5.

EditAgent produced about 26,000 training examples by keeping only verified results. Editors fine-tuned on this data performed better across all three benchmarks than those trained on data from FLUX. Adobe added AI agents to Photoshop, Premiere, and other Creative Cloud apps in June as a public beta. Black Forest Labs replaced the general-purpose model tested in the paper with Flux 3 Image Klein, which has 5.6 billion parameters.