Large multimodal models (LMMs) inherit the self-attention mechanism of pretrained language backbones, yet standard attention can exhibit suboptimal allocation, including cross-modal misallocation between textual and visual evidence and intra-visual imbalance among visual tokens. We propose RAVE (Re-Allocating Visual Attention), a lightweight pair-gating mechanism that adds a learned query-key bias to pre-softmax attention scores over visual keys, derived from pre-RoPE query and key features. RAVE requires no architectural modification to the backbone and can be trained end-to-end with the rest of the model. Across a suite of multimodal benchmarks, RAVE improves over standard attention by an average of 3 points, with the largest gains on perception-intensive tasks -- including multilingual OCR, chart understanding, document VQA, and scene text VQA -- where accurate visual grounding is critical.
@article{arxiv.2605.18359,
title = {RAVE: Re-Allocating Visual Attention in Large Multimodal Models},
author = {Xi Leng and Xinhong Ma and Ziqiang Dong and Feng Zhang and Xiaoying Tang and Yang Yang and Guanjun Jiang},
journal= {arXiv preprint arXiv:2605.18359},
year = {2026}
}