LoopVL: Recurrent Visual Intelligence

LoopVL, a recurrent vision-language model that combines Module-Loop and Model-Loop computation to iteratively update a unified vision-language state through shared modules, outperforms similarly sized and larger non-recurrent models on multimodal understanding and visual reasoning benchmarks, according to the arXiv paper 2609.38426v1. The authors report training LoopVL from scratch through language pre-training, multimodal training, and post-training, and observe "Visual Aha Moments" marked by pronounced shifts in visual attention across loops. The work offers practical evidence that shared parameters can support deeper multimodal computation over continuously evolving visual-language states.

arXiv:2609.38426v1 Announce Type: new Abstract: We introduce LoopVL to study whether Loop Transformers can be effectively extended to vision- language models. LoopVL combines Module-Loop and Model-Loop computation to iteratively update a unified vision-language state through shared modules. We train LoopVL from scratch through language pre-training, multimodal training, and post-training. LoopVL outperforms a range of similarly sized and larger non-recurrent models on multimodal understanding and visual reasoning benchmarks. We also observe Visual Aha Moments in LoopVL, characterized by pronounced shifts in visual attention across loops. LoopVL provides practical evidence for recurrent vision-language modeling and offers an intuitive perspective on how shared parameters can support deeper multimodal computation over continuously evolving visual-language states.