arXiv subjects

Jing Liu

Publications and source records attributed to Jing Liu.

At least 307 records · Page 17Linked to original sources

arXiv2024

Channel Merging: Preserving Specialization for Merged Experts

Lately, the practice of utilizing task-specific fine-tuning has been implemented to improve the performance of large language models (LLM) in subsequent tasks. Through the integration of diverse LLMs, the overall competency of LLMs is significantly boosted. Nevertheless, traditional ensemble methods are notably memory-intensive, necessitating the simultaneous loading of all specialized models into GPU memory. To address the inefficiency, model merging strategies have emerged, merging all LLMs into one model to reduce the memory footprint during inference. Despite these advances, model merging often leads to parameter conflicts and performance decline as the number of experts increases. Previous methods to mitigate these conflicts include post-pruning and partial merging. However, both approaches have limitations, particularly in terms of performance and storage efficiency when merged experts increase. To address these challenges, we introduce Channel Merging, a novel strategy designed to minimize parameter conflicts while enhancing storage efficiency. This method clusters and merges channel parameters based on their similarity to form several groups offline. By ensuring that only highly similar parameters are merged within each group, it significantly reduces parameter conflicts. During inference, we can instantly look up the expert parameters from the merged groups, preserving specialized knowledge. Our experiments demonstrate that Channel Merging consistently delivers high performance, matching unmerged models in tasks like English and Chinese reasoning, mathematical reasoning, and code generation. Moreover, it obtains results comparable to model ensemble with just 53% parameters when used with a task-specific router.

cs.CL↗

arXiv2024

Numerical Pruning for Efficient Autoregressive Models

Transformers have emerged as the leading architecture in deep learning, proving to be versatile and highly effective across diverse domains beyond language and image processing. However, their impressive performance often incurs high computational costs due to their substantial model size. This paper focuses on compressing decoder-only transformer-based autoregressive models through structural weight pruning to improve the model efficiency while preserving performance for both language and image generation tasks. Specifically, we propose a training-free pruning method that calculates a numerical score with Newton's method for the Attention and MLP modules, respectively. Besides, we further propose another compensation algorithm to recover the pruned model for better performance. To verify the effectiveness of our method, we provide both theoretical support and extensive experiments. Our experiments show that our method achieves state-of-the-art performance with reduced memory usage and faster generation speeds on GPUs.

cs.LG↗

arXiv2024

CA-SSLR: Condition-Aware Self-Supervised Learning Representation for Generalized Speech Processing

We introduce Condition-Aware Self-Supervised Learning Representation (CA-SSLR), a generalist conditioning model broadly applicable to various speech-processing tasks. Compared to standard fine-tuning methods that optimize for downstream models, CA-SSLR integrates language and speaker embeddings from earlier layers, making the SSL model aware of the current language and speaker context. This approach reduces the reliance on input audio features while preserving the integrity of the base SSLR. CA-SSLR improves the model's capabilities and demonstrates its generality on unseen tasks with minimal task-specific tuning. Our method employs linear modulation to dynamically adjust internal representations, enabling fine-grained adaptability without significantly altering the original model behavior. Experiments show that CA-SSLR reduces the number of trainable parameters, mitigates overfitting, and excels in under-resourced and unseen tasks. Specifically, CA-SSLR achieves a 10% relative reduction in LID errors, a 37% improvement in ASR CER on the ML-SUPERB benchmark, and a 27% decrease in SV EER on VoxCeleb-1, demonstrating its effectiveness.

eess.AS↗

arXiv2024

Detecting the dark sector through scalar-induced gravitational waves

We investigate the evolution of cosmological scalar perturbations in the case that the background radiation is weakly coupled to a light scalar field $ϕ$. The light scalar $ϕ$ is a homogeneous background field with a large initial value. In the radiation-dominated Universe, the coupling term introduces an effective mass to $ϕ$ and the background ultra-relativistic particles. The oscillations of $ϕ$ result in the periodic change of the equation of state parameter and the sound speed, which provides a novel mechanism to amplify subhorizon scalar perturbations through parametric resonance. The amplification of scalar perturbations leads to a stochastic gravitational-waves background~(SGWB) expected to be observed by multiband gravitational wave observers. The observation of the SGWB helps to determine the initial value of $ϕ$ and the coupling strength of the interaction. This mechanism is generally applicable to the interactions that introduce an effective mass, and we take the interaction between $ϕ$ and electrons as a concrete example to illustrate the result. We find that under the condition that the coupling coefficient $λ=10^{-16}$ and the initial value $ϕ_i=10^{18}$ GeV, the resulting SGWB spectrum is expected to be observed by the future observers including LISA, $Taiji$, DECIGO and BBO.

astro-ph.CO↗

arXiv2024

REAR: A Relevance-Aware Retrieval-Augmented Framework for Open-Domain Question Answering

Considering the limited internal parametric knowledge, retrieval-augmented generation (RAG) has been widely used to extend the knowledge scope of large language models (LLMs). Despite the extensive efforts on RAG research, in existing methods, LLMs cannot precisely assess the relevance of retrieved documents, thus likely leading to misleading or even incorrect utilization of external knowledge (eg., retrieved documents). To address this issue, in this paper, we propose REAR, a RElevance-Aware Retrieval-augmented approach for open-domain question answering (QA). As the key motivation, we aim to enhance the self-awareness regarding the reliability of external knowledge for LLMs, so as to adaptively utilize external knowledge in RAG systems. Specially, we develop a novel architecture for LLM-based RAG systems, by incorporating a specially designed assessment module that precisely assesses the relevance of retrieved documents. Furthermore, we propose an improved training method based on bi-granularity relevance fusion and noise-resistant training. By combining the improvements in both architecture and training, our proposed REAR can better utilize external knowledge by effectively perceiving the relevance of retrieved documents. Experiments on four open-domain QA tasks show that REAR significantly outperforms previous a number of competitive RAG approaches. Our codes can be accessed at https://github.com/RUCAIBox/REAR.

cs.CL↗

arXiv2024

Investigating the Factual Knowledge Boundary of Large Language Models with Retrieval Augmentation

Large language models (LLMs) have shown impressive prowess in solving a wide range of tasks with world knowledge. However, it remains unclear how well LLMs are able to perceive their factual knowledge boundaries, particularly under retrieval augmentation settings. In this study, we present the first analysis on the factual knowledge boundaries of LLMs and how retrieval augmentation affects LLMs on open-domain question answering (QA), with a bunch of important findings. Specifically, we focus on three research questions and analyze them by examining QA, priori judgement and posteriori judgement capabilities of LLMs. We show evidence that LLMs possess unwavering confidence in their knowledge and cannot handle the conflict between internal and external knowledge well. Furthermore, retrieval augmentation proves to be an effective approach in enhancing LLMs' awareness of knowledge boundaries. We further conduct thorough experiments to examine how different factors affect LLMs and propose a simple method to dynamically utilize supporting documents with our judgement strategy. Additionally, we find that the relevance between the supporting documents and the questions significantly impacts LLMs' QA and judgemental capabilities. The code to reproduce this work is available at https://github.com/RUCAIBox/LLM-Knowledge-Boundary.

cs.CL↗

arXiv2024

SynArtifact: Classifying and Alleviating Artifacts in Synthetic Images via Vision-Language Model

In the rapidly evolving area of image synthesis, a serious challenge is the presence of complex artifacts that compromise perceptual realism of synthetic images. To alleviate artifacts and improve quality of synthetic images, we fine-tune Vision-Language Model (VLM) as artifact classifier to automatically identify and classify a wide range of artifacts and provide supervision for further optimizing generative models. Specifically, we develop a comprehensive artifact taxonomy and construct a dataset of synthetic images with artifact annotations for fine-tuning VLM, named SynArtifact-1K. The fine-tuned VLM exhibits superior ability of identifying artifacts and outperforms the baseline by 25.66%. To our knowledge, this is the first time such end-to-end artifact classification task and solution have been proposed. Finally, we leverage the output of VLM as feedback to refine the generative model for alleviating artifacts. Visualization results and user study demonstrate that the quality of images synthesized by the refined diffusion model has been obviously improved.

cs.CV↗

arXiv2024

Self-Calibrated Listwise Reranking with Large Language Models

Large language models (LLMs), with advanced linguistic capabilities, have been employed in reranking tasks through a sequence-to-sequence approach. In this paradigm, multiple passages are reranked in a listwise manner and a textual reranked permutation is generated. However, due to the limited context window of LLMs, this reranking paradigm requires a sliding window strategy to iteratively handle larger candidate sets. This not only increases computational costs but also restricts the LLM from fully capturing all the comparison information for all candidates. To address these challenges, we propose a novel self-calibrated listwise reranking method, which aims to leverage LLMs to produce global relevance scores for ranking. To achieve it, we first propose the relevance-aware listwise reranking framework, which incorporates explicit list-view relevance scores to improve reranking efficiency and enable global comparison across the entire candidate set. Second, to ensure the comparability of the computed scores, we propose self-calibrated training that uses point-view relevance assessments generated internally by the LLM itself to calibrate the list-view relevance assessments. Extensive experiments and comprehensive analysis on the BEIR benchmark and TREC Deep Learning Tracks demonstrate the effectiveness and efficiency of our proposed method.

cs.IR↗

arXiv2024

Constraints on ultra-slow-roll inflation with the NANOGrav 15-Year Dataset

Ultra-slow-roll~(USR) inflation predicts an exponential amplification of scalar perturbations at small scales, which leads to a stochastic gravitational wave background~(SGWB) through the coupling of the scalar and tensor modes at the second-order expansion of the Einstein equation. In this work, we search for such a scalar-induced SGWB from the NANOGrav 15-year (NG15) dataset, and find that the SGWB from USR inflation could explain the observed data. The Bayes factors are $54\pm 5$ for the USR inflation model alone and $68\pm 6$ for the combined USR inflation plus supermassive black hole binaries (SMBHB) models. We place constraints on the amplitude of the scalar power spectrum to $P_{\mathrm{Rp}} > 10^{-1.95}$ at $95\%$ confidence level (C.L.) at the scale of $k\sim 20\, \mathrm{pc}^{-1}$. We find that $\log_{10} P_{\mathrm{Rp}}$ degenerates with the peak scale $\log_{10} k_{\mathrm{p}}$. We also obtain the parameter space allowed by the data in the USR inflationary scenario, where the $e$-folding numbers of the duration of the USR phase has a lower limit $ΔN > 2.80$ ($95\%$ C.L.) when the USR phase ends at $N\approx 20$. With astrophysically motivated priors, the NG15 dataset fits both the USR inflation model and SMBHB model equally well.

astro-ph.CO↗

arXiv2024

Quantum Diffusion Models for Few-Shot Learning

Modern quantum machine learning (QML) methods involve the variational optimization of parameterized quantum circuits on training datasets, followed by predictions on testing datasets. Most state-of-the-art QML algorithms currently lack practical advantages due to their limited learning capabilities, especially in few-shot learning tasks. In this work, we propose three new frameworks employing quantum diffusion model (QDM) as a solution for the few-shot learning: label-guided generation inference (LGGI); label-guided denoising inference (LGDI); and label-guided noise addition inference (LGNAI). Experimental results demonstrate that our proposed algorithms significantly outperform existing methods.

cs.LG↗

arXiv2024

SplatFace: Gaussian Splat Face Reconstruction Leveraging an Optimizable Surface

We present SplatFace, a novel Gaussian splatting framework designed for 3D human face reconstruction without reliance on accurate pre-determined geometry. Our method is designed to simultaneously deliver both high-quality novel view rendering and accurate 3D mesh reconstructions. We incorporate a generic 3D Morphable Model (3DMM) to provide a surface geometric structure, making it possible to reconstruct faces with a limited set of input images. We introduce a joint optimization strategy that refines both the Gaussians and the morphable surface through a synergistic non-rigid alignment process. A novel distance metric, splat-to-surface, is proposed to improve alignment by considering both the Gaussian position and covariance. The surface information is also utilized to incorporate a world-space densification process, resulting in superior reconstruction quality. Our experimental analysis demonstrates that the proposed method is competitive with both other Gaussian splatting techniques in novel view synthesis and other 3D reconstruction methods in producing 3D face meshes with high geometric precision.

cs.CV↗

arXiv2024

Rapid Plug-in Defenders

In the realm of daily services, the deployment of deep neural networks underscores the paramount importance of their reliability. However, the vulnerability of these networks to adversarial attacks, primarily evasion-based, poses a concerning threat to their functionality. Common methods for enhancing robustness involve heavy adversarial training or leveraging learned knowledge from clean data, both necessitating substantial computational resources. This inherent time-intensive nature severely limits the agility of large foundational models to swiftly counter adversarial perturbations. To address this challenge, this paper focuses on the Rapid Plug-in Defender (RaPiD) problem, aiming to rapidly counter adversarial perturbations without altering the deployed model. Drawing inspiration from the generalization and the universal computation ability of pre-trained transformer models, we propose a novel method termed CeTaD (Considering Pre-trained Transformers as Defenders) for RaPiD, optimized for efficient computation. CeTaD strategically fine-tunes the normalization layer parameters within the defender using a limited set of clean and adversarial examples. Our evaluation centers on assessing CeTaD's effectiveness, transferability, and the impact of different components in scenarios involving one-shot adversarial examples. The proposed method is capable of rapidly adapting to various attacks and different application scenarios without altering the target model and clean training data. We also explore the influence of varying training data conditions on CeTaD's performance. Notably, CeTaD exhibits adaptability across differentiable service models and proves the potential of continuous learning.

cs.CR↗

arXiv2024

Deep Insights into Automated Optimization with Large Language Models and Evolutionary Algorithms

Designing optimization approaches, whether heuristic or meta-heuristic, usually demands extensive manual intervention and has difficulty generalizing across diverse problem domains. The combination of Large Language Models (LLMs) and Evolutionary Algorithms (EAs) offers a promising new approach to overcome these limitations and make optimization more automated. In this setup, LLMs act as dynamic agents that can generate, refine, and interpret optimization strategies, while EAs efficiently explore complex solution spaces through evolutionary operators. Since this synergy enables a more efficient and creative search process, we first conduct an extensive review of recent research on the application of LLMs in optimization. We focus on LLMs' dual functionality as solution generators and algorithm designers. Then, we summarize the common and valuable designs in existing work and propose a novel LLM-EA paradigm for automated optimization. Furthermore, centered on this paradigm, we conduct an in-depth analysis of innovative methods for three key components: individual representation, variation operators, and fitness evaluation. We address challenges related to heuristic generation and solution exploration, especially from the LLM prompts' perspective. Our systematic review and thorough analysis of the paradigm can assist researchers in better understanding the current research and promoting the development of combining LLMs with EAs for automated optimization.

cs.NE↗

arXiv2024

ME-Switch: A Memory-Efficient Expert Switching Framework for Large Language Models

LLM development involves pre-training a foundation model on massive data, followed by fine-tuning on task-specific data to create specialized experts. Serving these experts can pose significant memory challenges, as loading all experts onto devices is impractical, and frequent switching between experts in response to user requests can incur substantial I/O costs. Previous approaches decompose the expert weights as the pre-trained weights plus delta weights, followed by quantizing the delta weights using output channel-wise step sizes to reduce the model size. However, these methods overlook the fact that certain input channels of delta weights can cause significant quantization errors at extremely low bitwidths. Additionally, existing methods assume that the appropriate model for a user request is known in advance, which is not the case in practice. To this end, we introduce ME-Switch, a memory-efficient expert switching framework tailored for serving multiple LLMs. To condense the number of bits required for describing the delta weights, we propose a salient-aware delta compression method that identifies salient input channels based on reconstruction error and applies mixed-precision quantization, reducing non-salient channels to low bits while keeping salient ones intact, cutting storage demand without compromising performance. Moreover, we develop a model-level routing method that efficiently directs user queries to the most suitable expert by performing domain classification. Extensive experiments show the promising memory efficiency and routing performance of ME-Switch. For example, when serving three models from the Mistral-7B family, ME-Switch reduces the model size by $1.74\times$ and maintains nearly lossless performance on instruction, mathematical reasoning, and code generation tasks. Notably, our method can efficiently serve 16 Mistral-7B models on a single NVIDIA A100 GPU.

cs.CL↗

arXiv2024

ChatSearch: a Dataset and a Generative Retrieval Model for General Conversational Image Retrieval

In this paper, we investigate the task of general conversational image retrieval on open-domain images. The objective is to search for images based on interactive conversations between humans and computers. To advance this task, we curate a dataset called ChatSearch. This dataset includes a multi-round multimodal conversational context query for each target image, thereby requiring the retrieval system to find the accurate image from database. Simultaneously, we propose a generative retrieval model named ChatSearcher, which is trained end-to-end to accept/produce interleaved image-text inputs/outputs. ChatSearcher exhibits strong capability in reasoning with multimodal context and can leverage world knowledge to yield visual retrieval results. It demonstrates superior performance on the ChatSearch dataset and also achieves competitive results on other image retrieval tasks and visual conversation tasks. We anticipate that this work will inspire further research on interactive multimodal retrieval systems. Our dataset will be available at https://github.com/joez17/ChatSearch.

cs.CV↗

arXiv2024

Liquid Metal Printed Superconducting Circuits

Since the discovery of superconductor one hundred years ago, tremendous theoretical and technological progresses have been achieved. The zero resistance and complete diamagnetism of superconducting materials promise many possibilities in diverse fields. However, the complexity and expensive manufacturing costs associated with the time-consuming superconductor fabrication process may retard their practices in a large extent. Here, via liquid metal printing we proposed to quickly fabricate superconducting electronics which can work at the prescribed cryogenic temperatures. By way of the room temperature fluidity of liquid metal composite inks, such one-step printing allows to pattern various superconducting circuits on the desired substrate. As the first-ever conceptual trial, the most easily available gallium-based liquid alloy inks were particularly adopted to composite with copper particles to achieve superconductivity under specific temperatures around 6.4K. Further, a series of liquid metal alloy and particles loaded composites were screened out and comparatively interpreted regarding their superconducting properties and potential values as printable inks in fabricating superconducting devices. The cost-effective feature and straightforward adaptability of the fabrication principle were evaluated. This work suggests an easy-going way for fabricating ending user superconducting devices, which may warrant more promising investigations and practices in the coming time.

cond-mat.supr-con↗

arXiv2024

AutoRNet: Automatically Optimizing Heuristics for Robust Network Design via Large Language Models

Achieving robust networks is a challenging problem due to its NP-hard nature and complex solution space. Current methods, from handcrafted feature extraction to deep learning, have made progress but remain rigid, requiring manual design and large labeled datasets. To address these issues, we propose AutoRNet, a framework that integrates large language models (LLMs) with evolutionary algorithms to generate heuristics for robust network design. We design network optimization strategies to provide domain-specific prompts for LLMs, utilizing domain knowledge to generate advanced heuristics. Additionally, we introduce an adaptive fitness function to balance convergence and diversity while maintaining degree distributions. AutoRNet is evaluated on sparse and dense scale-free networks, outperforming current methods by reducing the need for manual design and large datasets.

cs.AI↗

arXiv2024

Ada-K Routing: Boosting the Efficiency of MoE-based LLMs

In the era of Large Language Models (LLMs), Mixture-of-Experts (MoE) architectures offer a promising approach to managing computational costs while scaling up model parameters. Conventional MoE-based LLMs typically employ static Top-K routing, which activates a fixed and equal number of experts for each token regardless of their significance within the context. In this paper, we propose a novel Ada-K routing strategy that dynamically adjusts the number of activated experts for each token, thereby improving the balance between computational efficiency and model performance. Specifically, our strategy incorporates learnable and lightweight allocator modules that decide customized expert resource allocation tailored to the contextual needs for each token. These allocators are designed to be fully pluggable, making it broadly applicable across all mainstream MoE-based LLMs. We leverage the Proximal Policy Optimization (PPO) algorithm to facilitate an end-to-end learning process for this non-differentiable decision-making framework. Extensive evaluations on four popular baseline models demonstrate that our Ada-K routing method significantly outperforms conventional Top-K routing. Compared to Top-K, our method achieves over 25% reduction in FLOPs and more than 20% inference speedup while still improving performance across various benchmarks. Moreover, the training of Ada-K is highly efficient. Even for Mixtral-8x22B, a MoE-based LLM with more than 140B parameters, the training time is limited to 8 hours. Detailed analysis shows that harder tasks, middle layers, and content words tend to activate more experts, providing valuable insights for future adaptive MoE system designs. Both the training code and model checkpoints will be publicly available.

cs.CL↗