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.

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.

Tobuildabettergeneralized and easy-to-use pruning method, we propose AutoPrune, which prunes the network through optimizing a set of trainable auxiliary parameters instead of original weights.

Reducing the model redundancy is an important task to deploy complex deep learning models to resource-limited or time-sensitive devices. Directly regularizing or modifying weight values makes pruning procedure less robust and sensitive to the choice of hyperparameters, and it also requires prior knowledge to tune different hyperparameters for different models. To build a better generalized and easy-to-use pruning method, we propose AutoPrune, which prunes the network through optimizing a set of trainable auxiliary parameters instead of original weights. The instability and noise during training on auxiliary parameters will not directly affect weight values, which makes pruning process more robust to noise and less sensitive to hyperparameters. Moreover, we design gradient update rules for auxiliary parameters to keep them consistent with pruning tasks. Our method can automatically eliminate network redundancy with recoverability, relieving the complicated prior knowledge required to design thresholding functions, and reducing the time for trial and error. We evaluate our method with LeNet and VGG-like on MNIST and CIFAR-10 datasets, and with AlexNet, ResNet and MobileNet on ImageNet to establish the scalability of our work. Results show that our model achieves state-of-the-art sparsity, e.g.

After Rebuttal: The new test results are very welcome and indeed show that the proposed method is on-par with the most recent SOTA works in terms or compression rates, which greatly enhances its significance in my opinion. As a result I increased my score. While using masks is not entirely novel, having auxiliary parameters to decouple the pruning and learning processes is I think quite important and seems to be novel. While the lack of theoretical contributions makes the suitability of the paper to NeurIPS debatable, I think this still is a good paper and I wouldn't mind it being accepted. Since the paper focuses on pruning only this might be fine.

Reducing the model redundancy is an important task to deploy complex deep learning models to resource-limited or time-sensitive devices. Directly regularizing or modifying weight values makes pruning procedure less robust and sensitive to the choice of hyperparameters, and it also requires prior knowledge to tune different hyperparameters for different models. To build a better generalized and easy-to-use pruning method, we propose AutoPrune, which prunes the network through optimizing a set of trainable auxiliary parameters instead of original weights. The instability and noise during training on auxiliary parameters will not directly affect weight values, which makes pruning process more robust to noise and less sensitive to hyperparameters. Moreover, we design gradient update rules for auxiliary parameters to keep them consistent with pruning tasks.

Reducing the model redundancy is an important task to deploy complex deep learning models to resource-limited or time-sensitive devices. Directly regularizing or modifying weight values makes pruning procedure less robust and sensitive to the choice of hyperparameters, and it also requires prior knowledge to tune different hyperparameters for different models. To build a better generalized and easy-to-use pruning method, we propose AutoPrune, which prunes the network through optimizing a set of trainable auxiliary parameters instead of original weights. The instability and noise during training on auxiliary parameters will not directly affect weight values, which makes pruning process more robust to noise and less sensitive to hyperparameters. Moreover, we design gradient update rules for auxiliary parameters to keep them consistent with pruning tasks. Our method can automatically eliminate network redundancy with recoverability, relieving the complicated prior knowledge required to design thresholding functions, and reducing the time for trial and error.