ARCANA: A Reflective Multi-Agent Program Synthesis Framework for ARC-AGI-2 Reasoning
2026-07-10 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors created ARCANA, a system made up of multiple specialized agents that work together to solve complex reasoning tasks under tight time and hardware limits. ARCANA breaks down each problem into steps like seeing the scene, guessing a program to solve it, checking the program, and learning from mistakes. Each agent focuses on a different part, from understanding objects to testing solutions and improving based on errors. They all share information on a common platform and are guided by a controller that decides who acts when. This method helps solve difficult abstract tasks more efficiently and accurately.
multi-agent systemsymbolic executionscene graphdomain-specific languageprogram synthesisreflective refinementmeta controllerabstract reasoningblackboard architectureiterative perception
Authors
Kunbo Zhang, Lei Fu, Zeyu Wang, Zijing Liu, Kejian Tong
Abstract
We present ARCANA, a collaborative multi agent framework for solving ARC AGI 2 tasks under strict test time and hardware constraints. ARCANA decomposes each task into iterative perception, hypothesis generation, symbolic execution, and reflective refinement. A perceptual grounding agent builds object centric scene graphs from raw grids, a latent program policy proposes diverse DSL programs, a symbolic executor verifies candidates on demonstrations, and a reflective agent synthesizes failure driven feedback for the next turn. These agents communicate through a shared differentiable blackboard and are scheduled by a learned meta controller. The design combines structured program search with adaptive multi turn correction, improving reasoning efficiency and solution quality on challenging abstract transformation tasks.