FaSTA* agent uses LLMs and A* search for efficient multi-turn image editing

Researchers have developed FaSTA*, a neurosymbolic agent designed for efficient multi-turn image editing. This agent combines large language models for high-level task planning with A* search for detailed tool execution. To optimize costs, FaSTA* extracts and reuses common subroutines from successful toolpaths, enabling faster planning for recurring tasks and reserving the more computationally intensive A* search for novel challenges. The system demonstrates significant computational efficiency while maintaining competitive success rates compared to existing image editing methods. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces a cost-efficient agent for complex image editing tasks, potentially improving performance in creative and design applications.

RANK_REASON This is a research paper detailing a novel AI agent for image editing. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• FaSTA*
• arXiv
• A*
• LLM

• paper
• other