LoopViT: Scaling Visual ARC with Looped Transformers

TL;DR

LoopViT introduces a recursive vision transformer architecture with weight-tied recurrence and a dynamic exit mechanism, enabling more efficient and scalable visual reasoning that outperforms larger models on the ARC-AGI benchmark.

Contribution

It proposes LoopViT, a novel recursive transformer with weight tying and a dynamic halting mechanism, decoupling reasoning depth from model capacity for improved efficiency.

Findings

LoopViT achieves 65.8% accuracy on ARC-AGI-1 with only 18M parameters.

It outperforms larger 73M-parameter ensembles on the same benchmark.

Adaptive iterative computation proves more scalable than increasing network width.

Abstract

Recent advances in visual reasoning have leveraged vision transformers to tackle the ARC-AGI benchmark. However, we argue that the feed-forward architecture, where computational depth is strictly bound to parameter size, falls short of capturing the iterative, algorithmic nature of human induction. In this work, we propose a recursive architecture called Loop-ViT, which decouples reasoning depth from model capacity through weight-tied recurrence. Loop-ViT iterates a weight-tied Hybrid Block, combining local convolutions and global attention, to form a latent chain of thought. Crucially, we introduce a parameter-free Dynamic Exit mechanism based on predictive entropy: the model halts inference when its internal state ``crystallizes" into a low-uncertainty attractor. Empirical results on the ARC-AGI-1 benchmark validate this perspective: our 18M model achieves 65.8% accuracy, outperforming massive 73M-parameter ensembles. These findings demonstrate that adaptive iterative computation offers a far more efficient scaling axis for visual reasoning than simply increasing network width. The code is available at https://github.com/WenjieShu/LoopViT.