Jiaqi Nie

Yao Lu, Yutian Huang, Jiaqi Nie et al.

Recently, the field of machine learning has undergone a transition from model-centric to data-centric. The advancements in diverse learning tasks have been propelled by the accumulation of more extensive datasets, subsequently facilitating the training of larger models on these datasets. However, these datasets remain relatively under-explored. To this end, we introduce a pioneering approach known as RK-core, to empower gaining a deeper understanding of the intricate hierarchical structure within datasets. Across several benchmark datasets, we find that samples with low coreness values appear less representative of their respective categories, and conversely, those with high coreness values exhibit greater representativeness. Correspondingly, samples with high coreness values make a more substantial contribution to the performance in comparison to those with low coreness values. Building upon this, we further employ RK-core to analyze the hierarchical structure of samples with different coreset selection methods. Remarkably, we find that a high-quality coreset should exhibit hierarchical diversity instead of solely opting for representative samples. The code is available at https://github.com/yaolu-zjut/Kcore.

Shaoheng Wang, Yao Lu, Yuqi Li et al.

As a parameter efficient fine-tuning (PEFT) method, low-rank adaptation (LoRA) can save significant costs in storage and computing, but its strong adaptability to a single task is often accompanied by insufficient cross-task generalization capabilities. To improve this, existing work combines LoRA with mixture-of-experts (MoE) to enhance the model's adaptability through expert modules and routing mechanisms. However, existing LoRA-MoE methods lack unified standards in models, datasets, hyperparameters, and evaluation methods, making it difficult to conduct fair comparisons between different methods. To this end, we proposed a unified benchmark named LoRALib. Specifically, we standardized datasets from $40$ downstream tasks into a unified format, fine-tuned them using the same hyperparameters and obtained $680$ LoRA modules across $17$ model architectures. Based on this LoRA library, we conduct large-scale experiments on $3$ representative LoRA-MoE methods and different LoRA selection mechanisms using the open-sourced testing tool OpenCompass. Extensive experiments show that LoRAMoE performs best, and that prioritizing LoRAs relevant to the target task can further improve the performance of MoE. We hope these findings will inspire future work. Our datasets and LoRA library are available at https://huggingface.co/datasets/YaoLuzjut/LoRAOcean_dataset and https://huggingface.co/YaoLuzjut/models.

Yaxin Gao, Yao Lu, Zongfei Zhang et al.

Large language models (LLMs) have achieved remarkable success in many natural language processing (NLP) tasks. To achieve more accurate output, the prompts used to drive LLMs have become increasingly longer, which incurs higher computational costs. To address this prompt inflation problem, prompt compression has been proposed. However, most existing methods require training a small auxiliary model for compression, incurring a significant amount of additional computation. To avoid this, we propose a two-stage, training-free approach, called Dual-Stage Progressive Compression (DSPC). In the coarse-grained stage, semantic-related sentence filtering removes sentences with low semantic value based on TF-IDF. In the fine-grained stage, token importance is assessed using attention contribution, cross-model loss difference, and positional importance, enabling the pruning of low-utility tokens while preserving semantics. We validate DSPC on LLaMA-3.1-8B-Instruct and GPT-3.5-Turbo under a constrained token budget and observe consistent improvements. For instance, in the FewShot task of the Longbench dataset, DSPC achieves a performance of 49.17 by using only 3x fewer tokens, outperforming the best state-of-the-art baseline LongLLMLingua by 7.76.

Jack Cashman, Jiaqi Nie

The QMoE model provides a practical approach for compression of massive Mixture-of-Experts (MoE) models. QMoE offers a solution geared towards memory limitations that often reach terabyte scales, and it has the advantage of working with high sparsity models which implicitly lend themselves to compression techniques. QMoE also has the advantage of only taking MoE models into account and does not evaluate its use with non mixture of expert systems. Although this prior attempt focuses on the limitations of large servers with the latest NVIDIA hardware which in the case of the H100 and V100 which have 80 GB of HBM (High Bandwidth Memory), what is not being considered is a significantly more constrained environment, such as in the case of mobile devices which may have in the case of the iPhone anywhere from 4 to 8 GB of unified memory which also needs to be shared with the operating system and additional processes. Although edge devices such as phones and laptops are becoming increasingly more computationally powerful, they are still not close to the level of advanced server machines such as NVIDIA. An additional constraint that we must consider is that of latency. The communication time of sending a request to an LLM server and then getting it back is an additional waiting time that can be removed. We may also want to use LLM technology in environments where there is no reliable network connection.

Maoyu Wang, Yao Lu, Jiaqi Nie et al.

With the rapid development of deep learning, a growing number of pre-trained models have been publicly available. However, deploying these fixed models in real-world IoT applications is challenging because different devices possess heterogeneous computational and memory resources, making it impossible to deploy a single model across all platforms. Although traditional compression methods, such as pruning, quantization, and knowledge distillation, can improve efficiency, they become inflexible once applied and cannot adapt to changing resource constraints. To address these issues, we propose ReStNet, a Reusable and Stitchable Network that dynamically constructs a hybrid network by stitching two pre-trained models together. Implementing ReStNet requires addressing several key challenges, including how to select the optimal stitching points, determine the stitching order of the two pre-trained models, and choose an effective fine-tuning strategy. To systematically address these challenges and adapt to varying resource constraints, ReStNet determines the stitching point by calculating layer-wise similarity via Centered Kernel Alignment (CKA). It then constructs the hybrid model by retaining early layers from a larger-capacity model and appending deeper layers from a smaller one. To facilitate efficient deployment, only the stitching layer is fine-tuned. This design enables rapid adaptation to changing budgets while fully leveraging available resources. Moreover, ReStNet supports both homogeneous (CNN-CNN, Transformer-Transformer) and heterogeneous (CNN-Transformer) stitching, allowing to combine different model families flexibly. Extensive experiments on multiple benchmarks demonstrate that ReStNet achieve flexible accuracy-efficiency trade-offs at runtime while significantly reducing training cost.

Songtao Peng, Jiaqi Nie, Xincheng Shu et al.

As the default protocol for exchanging routing reachability information on the Internet, the abnormal behavior in traffic of Border Gateway Protocols (BGP) is closely related to Internet anomaly events. The BGP anomalous detection model ensures stable routing services on the Internet through its real-time monitoring and alerting capabilities. Previous studies either focused on the feature selection problem or the memory characteristic in data, while ignoring the relationship between features and the precise time correlation in feature (whether it's long or short term dependence). In this paper, we propose a multi-view model for capturing anomalous behaviors from BGP update traffic, in which Seasonal and Trend decomposition using Loess (STL) method is used to reduce the noise in the original time-series data, and Graph Attention Network (GAT) is used to discover feature relationships and time correlations in feature, respectively. Our results outperform the state-of-the-art methods at the anomaly detection task, with the average F1 score up to 96.3% and 93.2% on the balanced and imbalanced datasets respectively. Meanwhile, our model can be extended to classify multiple anomalous and to detect unknown events.