Feng Qian

Zhenguang Liu, Xinyang Yu, Ruili Wang et al.

The self-media era provides us tremendous high quality videos. Unfortunately, frequent video copyright infringements are now seriously damaging the interests and enthusiasm of video creators. Identifying infringing videos is therefore a compelling task. Current state-of-the-art methods tend to simply feed high-dimensional mixed video features into deep neural networks and count on the networks to extract useful representations. Despite its simplicity, this paradigm heavily relies on the original entangled features and lacks constraints guaranteeing that useful task-relevant semantics are extracted from the features. In this paper, we seek to tackle the above challenges from two aspects: (1) We propose to disentangle an original high-dimensional feature into multiple sub-features, explicitly disentangling the feature into exclusive lower-dimensional components. We expect the sub-features to encode non-overlapping semantics of the original feature and remove redundant information. (2) On top of the disentangled sub-features, we further learn an auxiliary feature to enhance the sub-features. We theoretically analyzed the mutual information between the label and the disentangled features, arriving at a loss that maximizes the extraction of task-relevant information from the original feature. Extensive experiments on two large-scale benchmark datasets (i.e., SVD and VCSL) demonstrate that our method achieves 90.1% TOP-100 mAP on the large-scale SVD dataset and also sets the new state-of-the-art on the VCSL benchmark dataset. Our code and model have been released at https://github.com/yyyooooo/DMI/, hoping to contribute to the community.

Yongji Wu, Xueshen Liu, Shuowei Jin et al.

The Mixture-of-Experts (MoE) architecture has become increasingly popular as a method to scale up large language models (LLMs). To save costs, heterogeneity-aware training solutions have been proposed to utilize GPU clusters made up of both newer and older-generation GPUs. However, existing solutions are agnostic to the performance characteristics of different MoE model components (i.e., attention and expert) and do not fully utilize each GPU's compute capability. In this paper, we introduce HeterMoE, a system to efficiently train MoE models on heterogeneous GPUs. Our key insight is that newer GPUs significantly outperform older generations on attention due to architectural advancements, while older GPUs are still relatively efficient for experts. HeterMoE disaggregates attention and expert computation, where older GPUs are only assigned with expert modules. Through the proposed zebra parallelism, HeterMoE overlaps the computation on different GPUs, in addition to employing an asymmetric expert assignment strategy for fine-grained load balancing to minimize GPU idle time. Our evaluation shows that HeterMoE achieves up to 2.3x speed-up compared to existing MoE training systems, and 1.4x compared to an optimally balanced heterogeneity-aware solution. HeterMoE efficiently utilizes older GPUs by maintaining 95% training throughput on average, even with half of the GPUs in a homogeneous A40 cluster replaced with V100.

Qian Feng, David S. Martinez Lema, Jianxiang Feng et al.

Learning dexterous manipulation from few-shot demonstrations is a significant yet challenging problem for advanced, human-like robotic systems. Dense distilled feature fields have addressed this challenge by distilling rich semantic features from 2D visual foundation models into the 3D domain. However, their reliance on neural rendering models such as Neural Radiance Fields (NeRF) or Gaussian Splatting results in high computational costs. In contrast, previous approaches based on sparse feature fields either suffer from inefficiencies due to multi-view dependencies and extensive training or lack sufficient grasp dexterity. To overcome these limitations, we propose Language-ENhanced Sparse Distilled Feature Field (LensDFF), which efficiently distills view-consistent 2D features onto 3D points using our novel language-enhanced feature fusion strategy, thereby enabling single-view few-shot generalization. Based on LensDFF, we further introduce a few-shot dexterous manipulation framework that integrates grasp primitives into the demonstrations to generate stable and highly dexterous grasps. Moreover, we present a real2sim grasp evaluation pipeline for efficient grasp assessment and hyperparameter tuning. Through extensive simulation experiments based on the real2sim pipeline and real-world experiments, our approach achieves competitive grasping performance, outperforming state-of-the-art approaches.