[2510.04871] Less is More: Recursive Reasoning with Tiny Networks

Researchers introduce Tiny Recursive Model (TRM), a simplified recursive reasoning approach that outperforms large language models on complex puzzle tasks like ARC-AGI, Sudoku, and Maze solving. Using only 7M parameters in a 2-layer network, TRM achieves 45% accuracy on ARC-AGI-1 and 8% on ARC-AGI-2, surpassing models like Show more