Adobe develops EditAgent for professional image editing
Original title: Adobe is testing how well AI agents can use editing tools like a Photoshop professional would
EditAgent generates parallel variants and uses verification data to outperform fine-tuned models on 600-edit benchmarks.
Adobe researchers built EditAgent, an AI agent that pairs general-purpose editors like Flux-2-Klein-9B with specialized tools for image editing tasks. The system generates multiple candidates per step and lets a language model judge them using close-up images, change maps, and statistics instead of relying on visual inspection alone. Tests show EditAgent beats seven other models including EditScore-72B, while also generating verified training data that improves open-source editors.
- General-purpose editor failed 42.7% of 600 test edits
- Professionals preferred parallel variants 78.5% of the time
- EditAgent produced 26,000 verified training examples
- Qwen3.8-27B with verification data scored 81.3% usable picks
- Vendors
- Adobe
- Organizations
- Adobe ResearchUC San Diego
- AI models
- EditAgentQwen3.8-27B
From the original article
4 min read at the-decoder.com
Adobe Research and UC San Diego built EditAgent, an AI system that edits images more like a Photoshop pro. Instead of retrying the same general-purpose editor over and over, the agent picks the right specialized tool for each task and lets a standard language model choose the best result based on measured image data. The approach works, but it needs about ten times the compute.
By Jonathan Kemper · the-decoder.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.