ASAP: Attention-Shift-Aware Pruning for Efficient LVLM Inference
While Large Vision-Language Models (LVLMs) demonstrate exceptional multi-modal capabilities, the quadratic computational cost of processing high-resolution visual tokens remains a critical bottleneck. Though recent token reduction strategies attempt to accelerate inference, such methods inadequately exploit attention values and fail to address token redundancy. More critically, they overlook the ``attention shift'' phenomenon inherent in LVLMs, which skews token attention scores. In this work, we propose ASAP, a novel training-free, KV-Cache-compatible pruning recipe that comprehensively addresses these limitations. First, we mitigate the attention shift by utilizing a dynamic bidirectional soft attention mask, ensuring the selection of genuinely informative tokens rather than naive attention-based selection. Second, we posit that high semantic redundancy within the token set degrades performance. We therefore introduce a weighted soft merging component that merges semantically similar tokens, preserving only the most feature-dense visual patches for subsequent layers. ASAP achieves virtually lossless compression of visual context, retaining 99.02% of the original LLaVA-NeXT-7B performance while aggressively slashing computational FLOPs by ~80%.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
ASAP: Attention Sink Anchored Pruning
Vision Transformers (ViTs) face severe computational bottlenecks due to the quadratic complexity of self-attention at high resolutions. Existing token reduction methods rely on local metrics - such as single-layer attent…
Adaptive Sharpness-Aware Pruning for Robust Sparse Networks
Robustness and compactness are two essential attributes of deep learning models that are deployed in the real world. The goals of robustness and compactness may seem to be at odds, since robustness requires generalizatio…
image-classificationImage Classificationobject-detectionObject Detection+1ASAP: Learning Generalizable Online Bin Packing via Adaptive Selection After Pruning
Recently, deep reinforcement learning (DRL) has achieved promising results in solving online 3D Bin Packing Problems (3D-BPP). However, these DRL-based policies may perform poorly on new instances due to distribution shi…
3D Bin PackingDecision MakingDeep Reinforcement LearningMeta-LearningDecoupled Similarity for Task-Aware Token Pruning in Large Vision-Language Models
Token pruning has emerged as an effective approach to reduce the substantial computational overhead of Large Vision-Language Models (LVLMs) by discarding less informative visual tokens while preserving performance. Howev…
ASAP: Unsupervised Post-training with Label Distribution Shift Adaptive Learning Rate
In real-world applications, machine learning models face online label shift, where label distributions change over time. Effective adaptation requires careful learning rate selection: too low slows adaptation and too hig…