Each Complexity Deserves a Pruning Policy

Ke Wang (The Chinese University of Hong Kong) · Zhipeng Zhang (Didi Research) · Weiming Hu (Institute of automation, Chinese academy of science) · Hanshi Wang (Institute of automation, Chinese academy of science) · Jin Gao (Chinese Academy of Sciences Institute of Automation) · Yufan Liu (Institute of automation, Chinese academy of science, Chinese Academy of Sciences) · Yuhao Xu (Sichuan University) · Zekun Xu (Institute of automation, Chinese academy of science, Chinese Academy of Sciences)

cognitive sciencecomplexity-adaptive pruningcomputational constraints empirical evidenceholistic reasoning trajectoryinference flopslarge vision-language modelslogistic retention curvemutual information pruning strategies sample complexity task complexity training-free framework vision-language-action models visual tokens

The established redundancy in visual tokens within large vision–language models (LVLMs) allows for pruning to effectively reduce their substantial computational demands. Empirical evidence from previous works indicates that visual tokens in later decoder stages receive less attention than shallow layers. Then, previous methods typically employ heuristics layer-specific pruning strategies where, although the number of tokens removed may differ across decoder layers, the overall pruning schedule is fixed and applied uniformly to all input samples and tasks, failing to align token elimination with the model’s holistic reasoning trajectory. Cognitive science indicates that human visual processing often begins with broad exploration to accumulate evidence before narrowing focus as the target becomes distinct. Our experiments reveal an analogous pattern in LVLMs. This observation strongly suggests that neither a fixed pruning schedule nor a heuristics layer-wise strategy can optimally accommodate the diverse complexities inherent in different inputs. To overcome this limitation, we introduce Complexity-Adaptive Pruning (AutoPrune), which is a training-free, plug-and-play framework that tailors pruning policies to varying sample and task complexities. Specifically, AutoPrune quantifies the mutual information between visual and textual tokens, and then projects this signal to a budget-constrained logistic retention curve. Each such logistic curve, defined by its unique shape, is shown to effectively correspond with the specific complexity of different tasks, and can easily guarantee adherence to a pre-defined computational constraints. We evaluate AutoPrune not only on standard vision-language tasks but also on Vision-Language-Action (VLA) models for autonomous driving. Notably, when applied to LLaVA-1.5-7B, our method prunes 89% of visual tokens and reduces inference FLOPs by 76.8%, but still retaining 96.7% of the original accuracy averaged over all tasks. This corresponds to a 9.1% improvement over the recent work PDrop (CVPR'2025), demonstrating the effectivenes. Code is available at https://github.com/AutoLab-SAI-SJTU/AutoPrune.