# CanvasAgent: Enabling Complex Image Creation and Editing via Visual Tool Orchestration

arXiv: https://arxiv.org/abs/2607.05465 · HTML: https://mlsanity.com/paper/2607.05465/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 6 of its 66 references that are in the map.
- First posted on arXiv 6 Jul 2026.

## Builds on (most cited first, 6 of 6)

- [Online Clustering of Bandits](https://mlsanity.com/paper/1401.8257/) (2014)
- [The Impact of Random Models on Clustering Similarity](https://mlsanity.com/paper/1701.06508/) (2017)
- [Fast Distributed Bandits for Online Recommendation Systems](https://mlsanity.com/paper/2007.08061/) (2020)
- [Unsupervised Heterogeneous Coupling Learning for Categorical Representation](https://mlsanity.com/paper/2007.10720/) (2020)
- [Graph Clustering Bandits for Recommendation](https://mlsanity.com/paper/1605.00596/) (2016)
- [Context-Aware Bandits](https://mlsanity.com/paper/1510.03164/) (2015)

## Built on by (most cited first, 1 of 1)

- [Agentic Visual Generation: From Generative Models to Agentic Control](https://mlsanity.com/paper/2609.06758/) (2026)

## Abstract (arXiv)

Complex image creation and editing often require more than a single generation or editing model. A user request may involve synthesizing images, localizing objects, segmenting regions, editing selected content, compositing intermediate assets, reading text, and enhancing the final result. Such tasks shift multimodal agents from perception-augmented reasoning to manipulation-centered visual creation, where tools must actively transform visual states rather than merely inspect them. However, existing multimodal tool-use agents are mostly optimized for perception, search, or domain-specific editing, and lack large-scale supervision for executable image-creation trajectories. In this paper, we introduce CanvasCraft, a large-scale multimodal tool-use dataset for complex image creation and editing, and CanvasAgent, a tool-augmented multimodal agent that learns to orchestrate heterogeneous visual tools through multi-turn interaction. CanvasCraft contains 140K fully annotated executable trajectories and 10K RL task specifications. CanvasAgent is first trained with SFT to learn executable reasoning-action trajectories, and is then optimized with GRPO using a hybrid reward that combines outcome- and process-level signals. During rollout, CanvasAgent inspects intermediate results, tracks visual assets, and adapts tool decisions to the evolving visual state. Experiments evaluate both final image quality and trajectory behavior, demonstrating the effectiveness of CanvasAgent and the proposed dataset for complex multi-tool image creation workflows.