Zhe Xu

Zhe Xu, Jiasheng Ye, Xiaoran Liu et al.

Recently, significant efforts have been devoted to enhancing the long-context capabilities of Large Language Models (LLMs), particularly in long-context reasoning. To facilitate this research, we propose \textbf{DetectiveQA}, a dataset specifically designed for narrative reasoning within long contexts. We leverage detective novels, averaging over 100k tokens, to create a dataset containing 1200 human-annotated questions in both Chinese and English, each paired with corresponding reference reasoning steps. Furthermore, we introduce a step-wise reasoning metric, which enhances the evaluation of LLMs' reasoning processes. We validate our approach and evaluate the mainstream LLMs, including GPT-4, Claude, and LLaMA, revealing persistent long-context reasoning challenges and demonstrating their evidence-retrieval challenges. Our findings offer valuable insights into the study of long-context reasoning and lay the base for more rigorous evaluations.

Zhe Xu, Ray C. C. Cheung

Although convolutional neural networks (CNNs) are now widely used in various computer vision applications, its huge resource demanding on parameter storage and computation makes the deployment on mobile and embedded devices difficult. Recently, binary convolutional neural networks are explored to help alleviate this issue by quantizing both weights and activations with only 1 single bit. However, there may exist a noticeable accuracy degradation when compared with full-precision models. In this paper, we propose an improved training approach towards compact binary CNNs with higher accuracy. Trainable scaling factors for both weights and activations are introduced to increase the value range. These scaling factors will be trained jointly with other parameters via backpropagation. Besides, a specific training algorithm is developed including tight approximation for derivative of discontinuous binarization function and $L_2$ regularization acting on weight scaling factors. With these improvements, the binary CNN achieves 92.3% accuracy on CIFAR-10 with VGG-Small network. On ImageNet, our method also obtains 46.1% top-1 accuracy with AlexNet and 54.2% with Resnet-18 surpassing previous works.

Zhe Xu, Biao Min, Ray C. C. Cheung

Scene background initialization allows the recovery of a clear image without foreground objects from a video sequence, which is generally the first step in many computer vision and video processing applications. The process may be strongly affected by some challenges such as illumination changes, foreground cluttering, intermittent movement, etc. In this paper, a robust background initialization approach based on superpixel motion detection is proposed. Both spatial and temporal characteristics of frames are adopted to effectively eliminate foreground objects. A subsequence with stable illumination condition is first selected for background estimation. Images are segmented into superpixels to preserve spatial texture information and foreground objects are eliminated by superpixel motion filtering process. A low-complexity density-based clustering is then performed to generate reliable background candidates for final background determination. The approach has been evaluated on SBMnet dataset and it achieves a performance superior or comparable to other state-of-the-art works with faster processing speed. Moreover, in those complex and dynamic categories, the algorithm produces the best results showing the robustness against very challenging scenarios.