arXiv subjects

Yuyao Sun

Publications and source records attributed to Yuyao Sun.

8 recordsLinked to original sources

arXiv2026

Learning to Predict Middle-Layer Attention in MLLMs for Visual Token Pruning

Multimodal large language models (MLLMs) achieve strong performance across diverse vision-language tasks, but their efficiency is limited by the cost of processing numerous visual tokens. Visual token pruning can reduce this cost, but requires accurate token importance estimates. Recent studies have demonstrated that text-to-vision attention from middle language model layers can effectively guide visual token pruning, typically using attention from a predefined middle layer to select the visual tokens to retain. Two problems therefore remain. First, our analysis shows that the layer whose attention is most responsive to the question varies substantially across samples, making a fixed layer suboptimal. Second, obtaining attention from the appropriate middle layer requires processing numerous visual tokens through several language model layers, by which point considerable computation has already been spent. To address both problems, we propose Middle-layer Attention Prediction (MAP), which uses Question Contrastive Teacher Selection to identify a sample-specific teacher layer by contrasting attention under the original and reference questions, and distills attention from the selected layer into a lightweight predictor that estimates visual token importance from multi-modal input features. During inference, MAP combines the predicted importance scores with a diversity criterion to prune visual tokens before the first language model layer. Thus, MAP requires no attention maps for pruning and remains compatible with existing inference acceleration techniques. Across ten benchmarks on LLaVA-NeXT-7B, MAP retains 97.5% of the unpruned model performance with only 5.56% of the visual tokens, yielding a 3.09x end-to-end speedup.

cs.AI↗

arXiv2025

InternVL-X: Advancing and Accelerating InternVL Series with Efficient Visual Token Compression

Most multimodal large language models (MLLMs) treat visual tokens as "a sequence of text", integrating them with text tokens into a large language model (LLM). However, a great quantity of visual tokens significantly increases the demand for computational resources and time. In this paper, we propose InternVL-X, which outperforms the InternVL model in both performance and efficiency by incorporating three visual token compression methods. First, we propose a novel vision-language projector, PVTC. This component integrates adjacent visual embeddings to form a local query and utilizes the transformed CLS token as a global query, then performs point-to-region cross-attention through these local and global queries to more effectively convert visual features. Second, we present a layer-wise visual token compression module, LVTC, which compresses tokens in the LLM shallow layers and then expands them through upsampling and residual connections in the deeper layers. This significantly enhances the model computational efficiency. Futhermore, we propose an efficient high resolution slicing method, RVTC, which dynamically adjusts the number of visual tokens based on image area or length filtering. RVTC greatly enhances training efficiency with only a slight reduction in performance. By utilizing 20% or fewer visual tokens, InternVL-X achieves state-of-the-art performance on 7 public MLLM benchmarks, and improves the average metric by 2.34% across 12 tasks.

cs.CV↗

arXiv2025

Towards Aligned Data Forgetting via Twin Machine Unlearning

Modern privacy regulations have spurred the evolution of machine unlearning, a technique enabling a trained model to efficiently forget specific training data. In prior unlearning methods, the concept of "data forgetting" is often interpreted and implemented as achieving zero classification accuracy on such data. Nevertheless, the authentic aim of machine unlearning is to achieve alignment between the unlearned model and the gold model, i.e., encouraging them to have identical classification accuracy. On the other hand, the gold model often exhibits non-zero classification accuracy due to its generalization ability. To achieve aligned data forgetting, we propose a Twin Machine Unlearning (TMU) approach, where a twin unlearning problem is defined corresponding to the original unlearning problem. Consequently, the generalization-label predictor trained on the twin problem can be transferred to the original problem, facilitating aligned data forgetting. Comprehensive empirical experiments illustrate that our approach significantly enhances the alignment between the unlearned model and the gold model.

cs.LG↗

arXiv2024

Towards Aligned Data Removal via Twin Machine Unlearning

Modern privacy regulations have spurred the evolution of machine unlearning, a technique that enables the removal of data from an already trained ML model without requiring retraining from scratch. Previous unlearning methods tend to induce the model to achieve lowest classification accuracy on the removal data. Nonetheless, the authentic objective of machine unlearning is to align the unlearned model with the gold model, i.e., achieving the same classification accuracy as the gold model. For this purpose, we present a Twin Machine Unlearning (TMU) approach, where a twin unlearning problem is defined corresponding to the original unlearning problem. As a results, the generalization-label predictor trained on the twin problem can be transferred to the original problem, facilitating aligned data removal. Comprehensive empirical experiments illustrate that our approach significantly enhances the alignment between the unlearned model and the gold model. Meanwhile, our method allows data removal without compromising the model accuracy.

cs.LG↗

arXiv2024

Towards Unified Robustness Against Both Backdoor and Adversarial Attacks

Deep Neural Networks (DNNs) are known to be vulnerable to both backdoor and adversarial attacks. In the literature, these two types of attacks are commonly treated as distinct robustness problems and solved separately, since they belong to training-time and inference-time attacks respectively. However, this paper revealed that there is an intriguing connection between them: (1) planting a backdoor into a model will significantly affect the model's adversarial examples; (2) for an infected model, its adversarial examples have similar features as the triggered images. Based on these observations, a novel Progressive Unified Defense (PUD) algorithm is proposed to defend against backdoor and adversarial attacks simultaneously. Specifically, our PUD has a progressive model purification scheme to jointly erase backdoors and enhance the model's adversarial robustness. At the early stage, the adversarial examples of infected models are utilized to erase backdoors. With the backdoor gradually erased, our model purification can naturally turn into a stage to boost the model's robustness against adversarial attacks. Besides, our PUD algorithm can effectively identify poisoned images, which allows the initial extra dataset not to be completely clean. Extensive experimental results show that, our discovered connection between backdoor and adversarial attacks is ubiquitous, no matter what type of backdoor attack. The proposed PUD outperforms the state-of-the-art backdoor defense, including the model repairing-based and data filtering-based methods. Besides, it also has the ability to compete with the most advanced adversarial defense methods.

cs.CV↗

arXiv2024

Active Admittance Control with Iterative Learning for General-Purpose Contact-Rich Manipulation

Force interaction is inevitable when robots face multiple operation scenarios. How to make the robot competent in force control for generalized operations such as multi-tasks still remains a challenging problem. Aiming at the reproducibility of interaction tasks and the lack of a generalized force control framework for multi-task scenarios, this paper proposes a novel hybrid control framework based on active admittance control with iterative learning parameters-tunning mechanism. The method adopts admittance control as the underlying algorithm to ensure flexibility, and iterative learning as the high-level algorithm to regulate the parameters of the admittance model. The whole algorithm has flexibility and learning ability, which is capable of achieving the goal of excellent versatility. Four representative interactive robot manipulation tasks are chosen to investigate the consistency and generalisability of the proposed method. Experiments are designed to verify the effectiveness of the whole framework, and an average of 98.21% and 91.52% improvement of RMSE is obtained relative to the traditional admittance control as well as the model-free adaptive control, respectively.

cs.RO↗

arXiv2021

A Lightweight Deep Network for Efficient CSI Feedback in Massive MIMO Systems

To fully exploit the advantages of massive multiple-input multiple-output (m-MIMO), accurate channel state information (CSI) is required at the transmitter. However, excessive CSI feedback for large antenna arrays is inefficient and thus undesirable in practical applications. By exploiting the inherent correlation characteristics of complex-valued channel responses in the angular-delay domain, we propose a novel neural network (NN) architecture, namely ENet, for CSI compression and feedback in m-MIMO. Even if the ENet processes the real and imaginary parts of the CSI values separately, its special structure enables the network trained for the real part only to be reused for the imaginary part. The proposed ENet shows enhanced performance with the network size reduced by nearly an order of magnitude compared to the existing NN-based solutions. Experimental results verify the effectiveness of the proposed ENet.

cs.IT↗

arXiv2020

AnciNet: An Efficient Deep Learning Approach for Feedback Compression of Estimated CSI in Massive MIMO Systems

Accurate channel state information (CSI) feedback plays a vital role in improving the performance gain of massive multiple-input multiple-output (m-MIMO) systems, where the dilemma is excessive CSI overhead versus limited feedback bandwith. By considering the noisy CSI due to imperfect channel estimation, we propose a novel deep neural network architecture, namely AnciNet, to conduct the CSI feedback with limited bandwidth. AnciNet extracts noise-free features from the noisy CSI samples to achieve effective CSI compression for the feedback. Experimental results verify that the proposed AnciNet approach outperforms the existing techniques under various conditions.

eess.SP↗