JarvisArt was released on Hugging Face It's an MLLM-driven agent for photo retouching, mimicking the reasoning of professional artists & coordinating 200+ tools in Adobe Lightroom. It even outperforms GPT-4o on content fidelity!
✨ Project Page: jarvisart.vercel.app 📄 Read the paper: huggingface.co/papers/2506.17… 💻 Code: github.com/LYL1015/Jarvis…
@HuggingPapers Our promotional video has been released, and the demo and preview weights will be released soon! jarvisart.vercel.app
@HuggingPapers @_akhaliq Only the paper has been published, code soonᵀᴹ
@HuggingPapers @_akhaliq This ties into AMD & mimik's edge-centric "ubiquitous execution." How will dense, local inference reshape agents?
@HuggingPapers @_akhaliq Impressive progress, but how might such agent tools function offline-executing adaptive, peer-orchestrated tasks on-device where connectivity is limited?