Multimodal Large Language Models (MLLMs) suffer from substantial computational overhead due to the high redundancy in visual token sequences. Existing approaches typically address this issue using single-layer Vision Transformer (ViT) features and static pruning strategies. However, such fixed configurations are often brittle under diverse instructions. To overcome these limitations, we propose CLASP, a plug-and-play token reduction framework based on class-adaptive layer fusion and dual-stage pruning. Specifically, CLASP first constructs category-specific visual representations through multi-layer vision feature fusion. It then performs dual-stage pruning, allocating the token budget between attention-salient pivot tokens for relevance and redundancy-aware completion tokens for coverage. Through class-adaptive pruning, CLASP enables prompt-conditioned feature fusion and budget allocation, allowing aggressive yet robust visual token reduction. Extensive experiments demonstrate that CLASP consistently outperforms existing methods across a wide range of benchmarks, pruning ratios, and MLLM architectures. Code will be available at https://github.com/Yunkaidang/CLASP.
@article{arxiv.2604.12767,
title = {CLASP: Class-Adaptive Layer Fusion and Dual-Stage Pruning for Multimodal Large Language Models},
author = {Yunkai Dang and Yizhu Jiang and Yifan Jiang and Qi Fan and Yinghuan Shi and Wenbin Li and Yang Gao},
journal= {arXiv preprint arXiv:2604.12767},
year = {2026}
}