On Twisted Roth-Lempel Codes
本文研究了一类扭曲的Roth-Lempel码，探讨了其最小距离、MDS和NMDS性质，并证明了这些码不同于传统的Roth-Lempel码。
0 citationsRead paper
Selected work
Representative Papers
Prototype Matters: Modality-unified Prototype Self-distillation for Unsupervised Visible-infrared Person Re-identification
本文提出一种新的跨模态学习框架，通过统一模态原型对比和原型引导的自蒸馏来优化模内和模间相似关系，以解决无监督可见-红外行人重识别中的可靠关联估计问题。
0 citationsRead paper
Fundamental Limits of Adaptive Beamforming Under Finite Training
本文探讨了有限训练对自适应波束形成输出SINR的影响，并通过Cramér-Rao界提供了一个下限，利用MVDR权重误差表达SINR损失。
0 citationsRead paper
Cut-ViT: Task-Specific Model Pruning via Gram Anchoring Subspace Consistency
为解决模型剪枝的鲁棒性和任务特异性问题，提出Cut-ViT方法，通过Gram锚定子空间一致性和谱熵适应进行高效剪枝。
0 citationsRead paper
Rethinking Visual Neglect: Steering via Context-Preference for MLLM Hallucination Mitigation
This work addresses the common misattribution of object hallucination in multimodal large language models to visual neglect. The authors propose a training-free, inference-time intervention framework that finely modulates the model’s reliance on visual context versus parametric knowledge through a contextual preference vector and a single-step residual injection mechanism applied at intermediate MLP layers. By integrating Contextual Preference Activation Steering (CAS) with carefully designed conflict samples, the method effectively mitigates object hallucinations without increasing decoding latency or compromising text generation quality, achieving a significant reduction in hallucination rates.
0 citationsRead paper
Recent publications
Latest Papers
On Twisted Roth-Lempel Codes
本文研究了一类扭曲的Roth-Lempel码，探讨了其最小距离、MDS和NMDS性质，并证明了这些码不同于传统的Roth-Lempel码。
0 citationsRead paper
Prototype Matters: Modality-unified Prototype Self-distillation for Unsupervised Visible-infrared Person Re-identification
本文提出一种新的跨模态学习框架，通过统一模态原型对比和原型引导的自蒸馏来优化模内和模间相似关系，以解决无监督可见-红外行人重识别中的可靠关联估计问题。
0 citationsRead paper
Fundamental Limits of Adaptive Beamforming Under Finite Training
本文探讨了有限训练对自适应波束形成输出SINR的影响，并通过Cramér-Rao界提供了一个下限，利用MVDR权重误差表达SINR损失。
0 citationsRead paper
Cut-ViT: Task-Specific Model Pruning via Gram Anchoring Subspace Consistency
为解决模型剪枝的鲁棒性和任务特异性问题，提出Cut-ViT方法，通过Gram锚定子空间一致性和谱熵适应进行高效剪枝。
0 citationsRead paper
Rethinking Visual Neglect: Steering via Context-Preference for MLLM Hallucination Mitigation
This work addresses the common misattribution of object hallucination in multimodal large language models to visual neglect. The authors propose a training-free, inference-time intervention framework that finely modulates the model’s reliance on visual context versus parametric knowledge through a contextual preference vector and a single-step residual injection mechanism applied at intermediate MLP layers. By integrating Contextual Preference Activation Steering (CAS) with carefully designed conflict samples, the method effectively mitigates object hallucinations without increasing decoding latency or compromising text generation quality, achieving a significant reduction in hallucination rates.
0 citationsRead paper