@mods creator inactive, and i think resolve? https://arcprize.org/blog/arc-prize-2025-results-analysis
According to this "Now in 2025, ARC-AGI is being used to demonstrate "refinement loops". From an information theory perspective, refinement is intelligence. While we still need new ideas to achieve AGI, ARC has catalyzed several now open-source refinement approaches (documented below). I anticipate these will push AI reasoning further in 2026."

So refinement loops were the most successful approach, particularly first place NVARC solution (paper here https://drive.google.com/file/d/1vkEluaaJTzaZiJL69TkZovJUkPSDH5Xc/view) which used a refinement loop for Multi-stage synthetic data generation.

its really vague but I think this meets the spirit of the question. Wouldnt be opposed to an N/A tho

@prismatic yeah having "Other" and "Other Approaches" kinda broke this market imo

Two Other categories? Unlinked binary so multiple could resolve? What's going on here? You need to be more descriptive in your criteria on how you plan to resolve this market.