Large multimodal models (LMMs) suffer significant computational challenges due to the high cost of Large Language Models (LLMs) and the quadratic complexity of processing long vision token sequences. In this paper, we explore the spatial redundancy among vision tokens and shorten the length of vision token sequences for inference acceleration. Specifically, we propose a Spatial Token Fusion (STF) method to learn compact vision tokens for short vision token sequence, where spatial-adjacent tokens are fused into one. Meanwhile, weight-frozen vision encoder can not well adapt to the demand of extensive downstream vision-language tasks. To this end, we further introduce a Multi-Block Token Fusion (MBTF) module to supplement multi-granularity features for the reduced token sequence. Overall, we combine STF and MBTF module to balance token reduction and information preservation, thereby improving inference efficiency without sacrificing multimodal reasoning capabilities. Experimental results demonstrate that our method based on LLaVA-1.5 achieves comparable or even superior performance to the baseline on 8 popular vision-language benchmarks with only $25\%$ vision tokens of baseline. The source code and trained weights are available at https://github.com/visresearch/LLaVA-STF.

Mamba and Vision Mamba (Vim) models have shown their potential as an alternative to methods based on Transformer architecture. This work introduces Fast Mamba for Vision (Famba-V), a cross-layer token fusion technique to enhance the training efficiency of Vim models. The key idea of Famba-V is to identify and fuse similar tokens across different Vim layers based on a suit of cross-layer strategies instead of simply applying token fusion uniformly across all the layers that existing works propose. We evaluate the performance of Famba-V on CIFAR-100. Our results show that Famba-V is able to enhance the training efficiency of Vim models by reducing both training time and peak memory usage during training. Moreover, the proposed cross-layer strategies allow Famba-V to deliver superior accuracy-efficiency trade-offs. These results all together demonstrate Famba-V as a promising efficiency enhancement technique for Vim models.

In this paper, we propose a method using the fusion of CNN and transformer structure to improve image classification performance. In the case of CNN, information about a local area on an image can be extracted well, but there is a limit to the extraction of global information. On the other hand, the transformer has an advantage in relatively global extraction, but has a disadvantage in that it requires a lot of memory for local feature value extraction. In the case of an image, it is converted into a feature map through CNN, and each feature map's pixel is considered a token. At the same time, the image is divided into patch areas and then fused with the transformer method that views them as tokens. For the fusion of tokens with two different characteristics, we propose three methods: (1) late token fusion with parallel structure, (2) early token fusion, (3) token fusion in a layer by layer. In an experiment using ImageNet 1k, the proposed method shows the best classification performance.