arXiv subjects

Han Zhang

Publications and source records attributed to Han Zhang.

At least 235 records · Page 13Linked to original sources

arXiv2024

COPR: Continual Human Preference Learning via Optimal Policy Regularization

Reinforcement Learning from Human Feedback (RLHF) is commonly utilized to improve the alignment of Large Language Models (LLMs) with human preferences. Given the evolving nature of human preferences, continual alignment becomes more crucial and practical in comparison to traditional static alignment. Nevertheless, making RLHF compatible with Continual Learning (CL) is challenging due to its complex process. Meanwhile, directly learning new human preferences may lead to Catastrophic Forgetting (CF) of historical preferences, resulting in helpless or harmful outputs. To overcome these challenges, we propose the Continual Optimal Policy Regularization (COPR) method, which draws inspiration from the optimal policy theory. COPR utilizes a sampling distribution as a demonstration and regularization constraints for CL. It adopts the Lagrangian Duality (LD) method to dynamically regularize the current policy based on the historically optimal policy, which prevents CF and avoids over-emphasizing unbalanced objectives. We also provide formal proof for the learnability of COPR. The experimental results show that COPR outperforms strong CL baselines on our proposed benchmark, in terms of reward-based, GPT-4 evaluations and human assessment. Furthermore, we validate the robustness of COPR under various CL settings, including different backbones, replay memory sizes, and learning orders.

cs.LG↗

arXiv2024

Correcting Large Language Model Behavior via Influence Function

Recent advancements in AI alignment techniques have significantly improved the alignment of large language models (LLMs) with static human preferences. However, the dynamic nature of human preferences can render some prior training data outdated or even erroneous, ultimately causing LLMs to deviate from contemporary human preferences and societal norms. Existing methodologies, whether they involve the curation of new data for continual alignment or the manual correction of outdated data for re-alignment, demand costly human resources. To address this challenge, we propose a novel approach, Large Language Model Behavior Correction with Influence Function Recall and Post-Training (LANCET), which requires no human involvement. LANCET consists of two phases: (1) using influence functions to identify the training data that significantly impact undesirable model outputs, and (2) applying an Influence function-driven Bregman Optimization (IBO) technique to adjust the model's behavior based on these influence distributions. Our experiments demonstrate that LANCET effectively and efficiently correct inappropriate behaviors of LLMs. Furthermore, LANCET can outperform methods that rely on collecting human preferences, and it enhances the interpretability of learning human preferences within LLMs.

cs.LG↗

arXiv2024

ChatDiT: A Training-Free Baseline for Task-Agnostic Free-Form Chatting with Diffusion Transformers

Recent research arXiv:2410.15027 arXiv:2410.23775 has highlighted the inherent in-context generation capabilities of pretrained diffusion transformers (DiTs), enabling them to seamlessly adapt to diverse visual tasks with minimal or no architectural modifications. These capabilities are unlocked by concatenating self-attention tokens across multiple input and target images, combined with grouped and masked generation pipelines. Building upon this foundation, we present ChatDiT, a zero-shot, general-purpose, and interactive visual generation framework that leverages pretrained diffusion transformers in their original form, requiring no additional tuning, adapters, or modifications. Users can interact with ChatDiT to create interleaved text-image articles, multi-page picture books, edit images, design IP derivatives, or develop character design settings, all through free-form natural language across one or more conversational rounds. At its core, ChatDiT employs a multi-agent system comprising three key components: an Instruction-Parsing agent that interprets user-uploaded images and instructions, a Strategy-Planning agent that devises single-step or multi-step generation actions, and an Execution agent that performs these actions using an in-context toolkit of diffusion transformers. We thoroughly evaluate ChatDiT on IDEA-Bench arXiv:2412.11767, comprising 100 real-world design tasks and 275 cases with diverse instructions and varying numbers of input and target images. Despite its simplicity and training-free approach, ChatDiT surpasses all competitors, including those specifically designed and trained on extensive multi-task datasets. We further identify key limitations of pretrained DiTs in zero-shot adapting to tasks. We release all code, agents, results, and intermediate outputs to facilitate further research at https://github.com/ali-vilab/ChatDiT

cs.CV↗

arXiv2024

Efficient Speech Command Recognition Leveraging Spiking Neural Network and Curriculum Learning-based Knowledge Distillation

The intrinsic dynamics and event-driven nature of spiking neural networks (SNNs) make them excel in processing temporal information by naturally utilizing embedded time sequences as time steps. Recent studies adopting this approach have demonstrated SNNs' effectiveness in speech command recognition, achieving high performance by employing large time steps for long time sequences. However, the large time steps lead to increased deployment burdens for edge computing applications. Thus, it is important to balance high performance and low energy consumption when detecting temporal patterns in edge devices. Our solution comprises two key components. 1). We propose a high-performance fully spike-driven framework termed SpikeSCR, characterized by a global-local hybrid structure for efficient representation learning, which exhibits long-term learning capabilities with extended time steps. 2). To further fully embrace low energy consumption, we propose an effective knowledge distillation method based on curriculum learning (KDCL), where valuable representations learned from the easy curriculum are progressively transferred to the hard curriculum with minor loss, striking a trade-off between power efficiency and high performance. We evaluate our method on three benchmark datasets: the Spiking Heidelberg Dataset (SHD), the Spiking Speech Commands (SSC), and the Google Speech Commands (GSC) V2. Our experimental results demonstrate that SpikeSCR outperforms current state-of-the-art (SOTA) methods across these three datasets with the same time steps. Furthermore, by executing KDCL, we reduce the number of time steps by 60% and decrease energy consumption by 54.8% while maintaining comparable performance to recent SOTA results. Therefore, this work offers valuable insights for tackling temporal processing challenges with long time sequences in edge neuromorphic computing systems.

cs.LG↗

arXiv2024

Lipschitz Singularities in Diffusion Models

Diffusion models, which employ stochastic differential equations to sample images through integrals, have emerged as a dominant class of generative models. However, the rationality of the diffusion process itself receives limited attention, leaving the question of whether the problem is well-posed and well-conditioned. In this paper, we explore a perplexing tendency of diffusion models: they often display the infinite Lipschitz property of the network with respect to time variable near the zero point. We provide theoretical proofs to illustrate the presence of infinite Lipschitz constants and empirical results to confirm it. The Lipschitz singularities pose a threat to the stability and accuracy during both the training and inference processes of diffusion models. Therefore, the mitigation of Lipschitz singularities holds great potential for enhancing the performance of diffusion models. To address this challenge, we propose a novel approach, dubbed E-TSDM, which alleviates the Lipschitz singularities of the diffusion model near the zero point of timesteps. Remarkably, our technique yields a substantial improvement in performance. Moreover, as a byproduct of our method, we achieve a dramatic reduction in the Fréchet Inception Distance of acceleration methods relying on network Lipschitz, including DDIM and DPM-Solver, by over 33%. Extensive experiments on diverse datasets validate our theory and method. Our work may advance the understanding of the general diffusion process, and also provide insights for the design of diffusion models.

cs.CV↗

arXiv2024

The Matrix: Infinite-Horizon World Generation with Real-Time Moving Control

We present The Matrix, the first foundational realistic world simulator capable of generating continuous 720p high-fidelity real-scene video streams with real-time, responsive control in both first- and third-person perspectives, enabling immersive exploration of richly dynamic environments. Trained on limited supervised data from AAA games like Forza Horizon 5 and Cyberpunk 2077, complemented by large-scale unsupervised footage from real-world settings like Tokyo streets, The Matrix allows users to traverse diverse terrains -- deserts, grasslands, water bodies, and urban landscapes -- in continuous, uncut hour-long sequences. Operating at 16 FPS, the system supports real-time interactivity and demonstrates zero-shot generalization, translating virtual game environments to real-world contexts where collecting continuous movement data is often infeasible. For example, The Matrix can simulate a BMW X3 driving through an office setting--an environment present in neither gaming data nor real-world sources. This approach showcases the potential of AAA game data to advance robust world models, bridging the gap between simulations and real-world applications in scenarios with limited data.

cs.AI↗

arXiv2024

Pan-protein Design Learning Enables Task-adaptive Generalization for Low-resource Enzyme Design

Computational protein design (CPD) offers transformative potential for bioengineering, but current deep CPD models, focused on universal domains, struggle with function-specific designs. This work introduces a novel CPD paradigm tailored for functional design tasks, particularly for enzymes-a key protein class often lacking specific application efficiency. To address structural data scarcity, we present CrossDesign, a domain-adaptive framework that leverages pretrained protein language models (PPLMs). By aligning protein structures with sequences, CrossDesign transfers pretrained knowledge to structure models, overcoming the limitations of limited structural data. The framework combines autoregressive (AR) and non-autoregressive (NAR) states in its encoder-decoder architecture, applying it to enzyme datasets and pan-proteins. Experimental results highlight CrossDesign's superior performance and robustness, especially with out-of-domain enzymes. Additionally, the model excels in fitness prediction when tested on large-scale mutation data, showcasing its stability.

q-bio.QM↗

arXiv2024

Compressive Spectrum Sensing with 1-bit ADCs

Efficient wideband spectrum sensing (WSS) is essential for managing spectrum scarcity in wireless communications. However, existing compressed sensing (CS)-based WSS methods require high sampling rates and power consumption, particularly with high-precision analog-to-digital converters (ADCs). Although 1-bit CS with low-precision ADCs can mitigate these demands, most approaches still depend on multi-user cooperation and prior sparsity information, which are often unavailable in WSS scenarios. This paper introduces a non-cooperative WSS method using multicoset sampling with 1-bit ADCs to achieve sub-Nyquist sampling without requiring sparsity knowledge. We analyze the impact of 1-bit quantization on multiband signals, then apply eigenvalue decomposition to isolate the signal subspace from noise, enabling spectrum support estimation without signal reconstruction. This approach provides a power-efficient solution for WSS that eliminates the need for cooperation and prior information.

eess.SP↗

arXiv2024

Voltage Support Capability Analysis of Grid-Forming Inverters with Current-Limiting Control Under Asymmetrical Grid Faults

Voltage support capability is critical for grid-forming (GFM) inverters with current-limiting control (CLC) during grid faults. Despite the findings on the voltage support for symmetrical grid faults, its applicability to more common but complex asymmetrical grid faults has yet to be verified rigorously. This letter fills the gap in the voltage support capability analysis for asymmetrical grid faults by establishing and analyzing positive- and negative-sequence equivalent circuit models, where the virtual impedance is adopted to emulate various CLCs. It is discovered that matching the phase angle of the virtual impedance, emulated by the CLC, with that of the composed impedance from the capacitor to the fault location can maximize the voltage support capability of GFM inverters under asymmetrical grid faults. Rigorous theoretical analysis and experimental results verify this conclusion.

eess.SY↗

arXiv2024

Cooperation and Personalization on a Seesaw: Choice-based FL for Safe Cooperation in Wireless Networks

Federated learning (FL) is an innovative distributed artificial intelligence (AI) technique. It has been used for interdisciplinary studies in different fields such as healthcare, marketing and finance. However the application of FL in wireless networks is still in its infancy. In this work, we first overview benefits and concerns when applying FL to wireless networks. Next, we provide a new perspective on existing personalized FL frameworks by analyzing the relationship between cooperation and personalization in these frameworks. Additionally, we discuss the possibility of tuning the cooperation level with a choice-based approach. Our choice-based FL approach is a flexible and safe FL framework that allows participants to lower the level of cooperation when they feel unsafe or unable to benefit from the cooperation. In this way, the choice-based FL framework aims to address the safety and fairness concerns in FL and protect participants from malicious attacks.

cs.NI↗

arXiv2024

ExaGRyPE: Numerical General Relativity Solvers Based upon the Hyperbolic PDEs Solver Engine ExaHyPE

ExaGRyPE describes a suite of solvers for numerical relativity, based upon ExaHyPE 2, the second generation of our Exascale Hyperbolic PDE Engine. The presented generation of ExaGRyPE solves the Einstein equations in the CCZ4 formulation under a 3+1 foliation and focuses on black hole spacetimes. It employs a block-structured Cartesian grid carrying a higher-order order Finite Difference scheme with adaptive mesh refinement, it facilitates massive parallelism combining message passing, domain decomposition and task parallelism, and it supports the injection of particles into the grid as data probes or tracers. We introduce the ExaGRyPE-specific building blocks within ExaHyPE 2, and discuss its software architecture and compute-n-feel: For this, we formalize the creation of any specific simulation with ExaGRyPE as a sequence of lowering operations, where abstract logical tasks are successively broken into finer tasks until we obtain an abstraction level that can be mapped onto a C++ executable. The overall program logic is fully specified via a domain-specific Python interface, we map this logic onto a more detailed set of numerical tasks, subsequently lower this representation onto technical tasks that the underlying ExaHyPE engine uses to parallelize the application, before eventually the technical tasks are mapped onto task graphs including the actual PDE term evaluations, initial conditions, boundary conditions, and so forth. We end up with a rigorous separation of concerns which shields ExaGRyPE users from technical details and hence simplifies the development of novel physical models. We present the simulations and data for the gauge wave, static single black holes and rotating binary black hole systems, demonstrating that the code base is mature and usable. However, we also uncover domain-specific numerical challenges that need further study by the community in future work.

gr-qc↗

arXiv2024

QIS : Interactive Segmentation via Quasi-Conformal Mappings

Image segmentation plays a crucial role in extracting important objects of interest from images, enabling various applications. While existing methods have shown success in segmenting clean images, they often struggle to produce accurate segmentation results when dealing with degraded images, such as those containing noise or occlusions. To address this challenge, interactive segmentation has emerged as a promising approach, allowing users to provide meaningful input to guide the segmentation process. However, an important problem in interactive segmentation lies in determining how to incorporate minimal yet meaningful user guidance into the segmentation model. In this paper, we propose the quasi-conformal interactive segmentation (QIS) model, which incorporates user input in the form of positive and negative clicks. Users mark a few pixels belonging to the object region as positive clicks, indicating that the segmentation model should include a region around these clicks. Conversely, negative clicks are provided on pixels belonging to the background, instructing the model to exclude the region near these clicks from the segmentation mask. Additionally, the segmentation mask is obtained by deforming a template mask with the same topology as the object of interest using an orientation-preserving quasiconformal mapping. This approach helps to avoid topological errors in the segmentation results. We provide a thorough analysis of the proposed model, including theoretical support for the ability of QIS to include or exclude regions of interest or disinterest based on the user's indication. To evaluate the performance of QIS, we conduct experiments on synthesized images, medical images, natural images and noisy natural images. The results demonstrate the efficacy of our proposed method.

cs.CV↗

arXiv2024

Hierarchical Knowledge Graph Construction from Images for Scalable E-Commerce

Knowledge Graph (KG) is playing an increasingly important role in various AI systems. For e-commerce, an efficient and low-cost automated knowledge graph construction method is the foundation of enabling various successful downstream applications. In this paper, we propose a novel method for constructing structured product knowledge graphs from raw product images. The method cooperatively leverages recent advances in the vision-language model (VLM) and large language model (LLM), fully automating the process and allowing timely graph updates. We also present a human-annotated e-commerce product dataset for benchmarking product property extraction in knowledge graph construction. Our method outperforms our baseline in all metrics and evaluated properties, demonstrating its effectiveness and bright usage potential.

cs.AI↗

arXiv2024

Multi-Modal Transformer and Reinforcement Learning-based Beam Management

Beam management is an important technique to improve signal strength and reduce interference in wireless communication systems. Recently, there has been increasing interest in using diverse sensing modalities for beam management. However, it remains a big challenge to process multi-modal data efficiently and extract useful information. On the other hand, the recently emerging multi-modal transformer (MMT) is a promising technique that can process multi-modal data by capturing long-range dependencies. While MMT is highly effective in handling multi-modal data and providing robust beam management, integrating reinforcement learning (RL) further enhances their adaptability in dynamic environments. In this work, we propose a two-step beam management method by combining MMT with RL for dynamic beam index prediction. In the first step, we divide available beam indices into several groups and leverage MMT to process diverse data modalities to predict the optimal beam group. In the second step, we employ RL for fast beam decision-making within each group, which in return maximizes throughput. Our proposed framework is tested on a 6G dataset. In this testing scenario, it achieves higher beam prediction accuracy and system throughput compared to both the MMT-only based method and the RL-only based method.

eess.SP↗

arXiv2024

QKFormer: Hierarchical Spiking Transformer using Q-K Attention

Spiking Transformers, which integrate Spiking Neural Networks (SNNs) with Transformer architectures, have attracted significant attention due to their potential for energy efficiency and high performance. However, existing models in this domain still suffer from suboptimal performance. We introduce several innovations to improve the performance: i) We propose a novel spike-form Q-K attention mechanism, tailored for SNNs, which efficiently models the importance of token or channel dimensions through binary vectors with linear complexity. ii) We incorporate the hierarchical structure, which significantly benefits the performance of both the brain and artificial neural networks, into spiking transformers to obtain multi-scale spiking representation. iii) We design a versatile and powerful patch embedding module with a deformed shortcut specifically for spiking transformers. Together, we develop QKFormer, a hierarchical spiking transformer based on Q-K attention with direct training. QKFormer shows significantly superior performance over existing state-of-the-art SNN models on various mainstream datasets. Notably, with comparable size to Spikformer (66.34 M, 74.81%), QKFormer (64.96 M) achieves a groundbreaking top-1 accuracy of 85.65% on ImageNet-1k, substantially outperforming Spikformer by 10.84%. To our best knowledge, this is the first time that directly training SNNs have exceeded 85% accuracy on ImageNet-1K. The code and models are publicly available at https://github.com/zhouchenlin2096/QKFormer

cs.NE↗

arXiv2024

Five-body $D\to V$ Semileptonic Decays

Our main objective is to derive the decay rate for the semileptonic decays $D\to V\ell^+ν_{\ell}\,(\ell=e,μ)$, where $V$ represents a vector particle. In these decays, the vector particle $V$ decays into three pseudo-scalar particles. To accomplish this, we evaluate the phase-space factor for the five-body decay with a set of eight independent variables which uniquely define a point in the phase space. We further conduct a detailed investigation of the $D\to ω\ell^+ν_{\ell}$, where $ω$ subsequently decays into $π^+π^-π^0$, within the Standard Model and in a general effective field theory description of the weak interactions at low energies. The outcomes of this study have potential applications in the measurement of $D\to ω$ form factors. These measurements can be performed using data obtained from BESIII.

hep-ph↗

arXiv2024

StraightTrack: Towards Mixed Reality Navigation System for Percutaneous K-wire Insertion

In percutaneous pelvic trauma surgery, accurate placement of Kirschner wires (K-wires) is crucial to ensure effective fracture fixation and avoid complications due to breaching the cortical bone along an unsuitable trajectory. Surgical navigation via mixed reality (MR) can help achieve precise wire placement in a low-profile form factor. Current approaches in this domain are as yet unsuitable for real-world deployment because they fall short of guaranteeing accurate visual feedback due to uncontrolled bending of the wire. To ensure accurate feedback, we introduce StraightTrack, an MR navigation system designed for percutaneous wire placement in complex anatomy. StraightTrack features a marker body equipped with a rigid access cannula that mitigates wire bending due to interactions with soft tissue and a covered bony surface. Integrated with an Optical See-Through Head-Mounted Display (OST HMD) capable of tracking the cannula body, StraightTrack offers real-time 3D visualization and guidance without external trackers, which are prone to losing line-of-sight. In phantom experiments with two experienced orthopedic surgeons, StraightTrack improves wire placement accuracy, achieving the ideal trajectory within $5.26 \pm 2.29$ mm and $2.88 \pm 1.49$ degree, compared to over 12.08 mm and 4.07 degree for comparable methods. As MR navigation systems continue to mature, StraightTrack realizes their potential for internal fracture fixation and other percutaneous orthopedic procedures.

cs.RO↗

arXiv2024

Understanding the Benefits of SimCLR Pre-Training in Two-Layer Convolutional Neural Networks

SimCLR is one of the most popular contrastive learning methods for vision tasks. It pre-trains deep neural networks based on a large amount of unlabeled data by teaching the model to distinguish between positive and negative pairs of augmented images. It is believed that SimCLR can pre-train a deep neural network to learn efficient representations that can lead to a better performance of future supervised fine-tuning. Despite its effectiveness, our theoretical understanding of the underlying mechanisms of SimCLR is still limited. In this paper, we theoretically introduce a case study of the SimCLR method. Specifically, we consider training a two-layer convolutional neural network (CNN) to learn a toy image data model. We show that, under certain conditions on the number of labeled data, SimCLR pre-training combined with supervised fine-tuning achieves almost optimal test loss. Notably, the label complexity for SimCLR pre-training is far less demanding compared to direct training on supervised data. Our analysis sheds light on the benefits of SimCLR in learning with fewer labels.

cs.LG↗