Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2508.06084 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914232168087552 |
|---|---|
| author | Zhang, Weichen Zhu, Zhui Li, Ningbo Tao, Shilong Liu, Kebin Liu, Yunhao |
| author_facet | Zhang, Weichen Zhu, Zhui Li, Ningbo Tao, Shilong Liu, Kebin Liu, Yunhao |
| contents | Vision-language models (VLMs) have achieved impressive performance on multimodal reasoning tasks such as visual question answering, image captioning and so on, but their inference cost remains a significant challenge due to the large number of vision tokens processed during the prefill stage. Existing pruning methods often rely on directly using the attention patterns or static text prompt guidance, failing to exploit the dynamic internal signals generated during inference. To address these issues, we propose AdaptInfer, a plug-and-play framework for adaptive vision token pruning in VLMs. First, we introduce a fine-grained, dynamic text-guided pruning mechanism that reuses layer-wise text-to-text attention maps to construct soft priors over text-token importance, allowing more informed scoring of vision tokens at each stage. Second, we perform an offline analysis of cross-modal attention shifts and identify consistent inflection locations in inference, which inspire us to propose a more principled and efficient pruning schedule. Our method is lightweight and plug-and-play, also generalizable across multi-modal tasks. Experimental results have verified the effectiveness of the proposed method. For example, it reduces CUDA latency by 61.3% while maintaining an average accuracy of 93.1% on vanilla LLaVA-1.5-7B. Under the same token budget, AdaptInfer surpasses SOTA in accuracy. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_06084 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | AdaptInfer: Adaptive Token Pruning for Vision-Language Model Inference with Dynamical Text Guidance Zhang, Weichen Zhu, Zhui Li, Ningbo Tao, Shilong Liu, Kebin Liu, Yunhao Computer Vision and Pattern Recognition Vision-language models (VLMs) have achieved impressive performance on multimodal reasoning tasks such as visual question answering, image captioning and so on, but their inference cost remains a significant challenge due to the large number of vision tokens processed during the prefill stage. Existing pruning methods often rely on directly using the attention patterns or static text prompt guidance, failing to exploit the dynamic internal signals generated during inference. To address these issues, we propose AdaptInfer, a plug-and-play framework for adaptive vision token pruning in VLMs. First, we introduce a fine-grained, dynamic text-guided pruning mechanism that reuses layer-wise text-to-text attention maps to construct soft priors over text-token importance, allowing more informed scoring of vision tokens at each stage. Second, we perform an offline analysis of cross-modal attention shifts and identify consistent inflection locations in inference, which inspire us to propose a more principled and efficient pruning schedule. Our method is lightweight and plug-and-play, also generalizable across multi-modal tasks. Experimental results have verified the effectiveness of the proposed method. For example, it reduces CUDA latency by 61.3% while maintaining an average accuracy of 93.1% on vanilla LLaVA-1.5-7B. Under the same token budget, AdaptInfer surpasses SOTA in accuracy. |
| title | AdaptInfer: Adaptive Token Pruning for Vision-Language Model Inference with Dynamical Text Guidance |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2508.06084 |