arXiv subjects

Lei Huang

Publications and source records attributed to Lei Huang.

At least 91 records · Page 5Linked to original sources

arXiv2025

An Investigation of Batch Normalization in Off-Policy Actor-Critic Algorithms

Batch Normalization (BN) has played a pivotal role in the success of deep learning by improving training stability, mitigating overfitting, and enabling more effective optimization. However, its adoption in deep reinforcement learning (DRL) has been limited due to the inherent non-i.i.d. nature of data and the dynamically shifting distributions induced by the agent's learning process. In this paper, we argue that, despite these challenges, BN retains unique advantages in DRL settings, particularly through its stochasticity and its ability to ease training. When applied appropriately, BN can adapt to evolving data distributions and enhance both convergence speed and final performance. To this end, we conduct a comprehensive empirical study on the use of BN in off-policy actor-critic algorithms, systematically analyzing how different training and evaluation modes impact performance. We further identify failure modes that lead to instability or divergence, analyze their underlying causes, and propose the Mode-Aware Batch Normalization (MA-BN) method with practical actionable recommendations for robust BN integration in DRL pipelines. We also empirically validate that, in RL settings, MA-BN accelerates and stabilizes training, broadens the effective learning rate range, enhances exploration, and reduces overall optimization difficulty. Our code is available at: https://github.com/monster476/ma-bn.git.

cs.LG↗

arXiv2025

Taming Variability: Randomized and Bootstrapped Conformal Risk Control for LLMs

We transform the randomness of LLMs into precise assurances using an actuator at the API interface that applies a user-defined risk constraint in finite samples via Conformal Risk Control (CRC). This label-free and model-agnostic actuator manages ship/abstain/escalate actions based solely on a scalar score from opaque outputs. We enhance CRC's computational efficiency and robustness through Batched Bootstrap CRC (BB-CRC) and Randomized Batched Weighted-Average CRC (RBWA-CRC), reducing calibration calls and stabilizing thresholds while maintaining statistical validity. Additionally, we present a semantic quantification method grounded in gram matrix geometry, resulting in interpretable signal and metric design. Together these pieces deliver principled randomness control for LLM hallucination mitigation and LLM-as-judge reliability. Our framework is assessed using four datasets, demonstrating its efficacy in enhancing factual accuracy and measuring LLM-as-judge performance, yielding a simplified and computationally efficient control layer that converts variability into statistical validity.

stat.ME↗

arXiv2025

Unsupervised Conformal Inference: Bootstrapping and Alignment to Control LLM Uncertainty

Deploying black-box LLMs requires managing uncertainty in the absence of token-level probability or true labels. We propose introducing an unsupervised conformal inference framework for generation, which integrates: generative models, incorporating: (i) an LLM-compatible atypical score derived from response-embedding Gram matrix, (ii) UCP combined with a bootstrapping variant (BB-UCP) that aggregates residuals to refine quantile precision while maintaining distribution-free, finite-sample coverage, and (iii) conformal alignment, which calibrates a single strictness parameter $τ$ so a user predicate (e.g., factuality lift) holds on unseen batches with probability $\ge 1-α$. Across different benchmark datasets, our gates achieve close-to-nominal coverage and provide tighter, more stable thresholds than split UCP, while consistently reducing the severity of hallucination, outperforming lightweight per-response detectors with similar computational demands. The result is a label-free, API-compatible gate for test-time filtering that turns geometric signals into calibrated, goal-aligned decisions.

stat.ML↗

arXiv2025

Context-Aware Hierarchical Taxonomy Generation for Scientific Papers via LLM-Guided Multi-Aspect Clustering

The rapid growth of scientific literature demands efficient methods to organize and synthesize research findings. Existing taxonomy construction methods, leveraging unsupervised clustering or direct prompting of large language models (LLMs), often lack coherence and granularity. We propose a novel context-aware hierarchical taxonomy generation framework that integrates LLM-guided multi-aspect encoding with dynamic clustering. Our method leverages LLMs to identify key aspects of each paper (e.g., methodology, dataset, evaluation) and generates aspect-specific paper summaries, which are then encoded and clustered along each aspect to form a coherent hierarchy. In addition, we introduce a new evaluation benchmark of 156 expert-crafted taxonomies encompassing 11.6k papers, providing the first naturally annotated dataset for this task. Experimental results demonstrate that our method significantly outperforms prior approaches, achieving state-of-the-art performance in taxonomy coherence, granularity, and interpretability.

cs.CL↗

arXiv2025

Bounds and Constructions of High-Memory Spatially-Coupled Codes

In this paper, we apply the Clique Lovász Local Lemma to provide sufficient conditions on memory and lifting degree for removing certain harmful combinatorial structures in spatially-coupled (SC) codes that negatively impact decoding performance. Additionally, we present, for the first time, a constructive algorithm based on the Moser-Tardos algorithm that ensures predictable performance. Furthermore, leveraging the properties of LLL-distribution and M-T-distribution, we establish the dependencies among the harmful structures during the construction process. We provide upper bounds on the probability change of remaining harmful structures after eliminating some of them. In particular, the elimination of 4-cycles increases the probability of 6-cycles becoming active by at most a factor of $e^{8/3}$.

cs.IT↗

arXiv2025

SAGE: Semantic-Aware Shared Sampling for Efficient Diffusion

Diffusion models manifest evident benefits across diverse domains, yet their high sampling cost, requiring dozens of sequential model evaluations, remains a major limitation. Prior efforts mainly accelerate sampling via optimized solvers or distillation, which treat each query independently. In contrast, we reduce total number of steps by sharing early-stage sampling across semantically similar queries. To enable such efficiency gains without sacrificing quality, we propose SAGE, a semantic-aware shared sampling framework that integrates a shared sampling scheme for efficiency and a tailored training strategy for quality preservation. Extensive experiments show that SAGE reduces sampling cost by 25.5%, while improving generation quality with 5.0% lower FID, 5.4% higher CLIP, and 160% higher diversity over baselines.

cs.LG↗

arXiv2025

Hierarchical Equations of Motion Solved with the Multiconfigurational Ehrenfest Ansatz

Being a numerically exact method for the simulation of dynamics in open quantum systems, the hierarchical equations of motion (HEOM) still suffers from the curse of dimensionality. In this study, we propose a novel MCE-HEOM method, which introduces the multiconfigurational Ehrenfest (MCE) ansatz to the second quantization formalism of HEOM. Here, the MCE equations of motion are derived from the time-dependent variational principle in a composed Hilbert-Liouville space, and each MCE coherent-state basis can be regarded as having an infinite hierarchical tier, such that the truncation tier of auxiliary density operators in MCE-HEOM can also be considered to be infinite. As demonstrated in a series of representative spin-boson models, our MCE-HEOM significantly reduces the number of variational parameters and could efficiently handle the strong non-Markovian effect, which is difficult for conventional HEOM due to the requirement of a very deep truncation tier. Compared with MCE, MCE-HEOM reduces the number of effective bath modes and circumvents the initial samplings for finite temperature, eventually resulting in a huge reduction of computational cost.

physics.chem-ph↗

arXiv2025

Nonreciprocal Entanglement by Dynamically Encircling a Nexus

Nonreciprocal entanglement, characterized by inherently robust operation, is a cornerstone for quantum information processing and communications. However, it remains a great challenge to achieve nonreciprocal entanglement characterized by stability and robustness against environmental fluctuations. Here, we propose a universal nonlinear mechanism to engineer magnetic-free nonreciprocity in dissipative optomechanics by utilizing bistability, a phenomenon ubiquitous across nonlinear physical systems. By dynamically encircling the nexus of bistability, a cusp converged by the bistable surfaces, we obtain nonreciprocal displacement and then utilize it to achieve robust nonreciprocal entanglement. Owing to the unique landscape of bistability, our nonreciprocal displacement and entanglements exhibit stability and robustness through closed-loop operations. Our work presents a foundational framework for leveraging nonlinearity to achieve nonreciprocal quantum information processing. It paves new avenues for exploring nonreciprocal quantum information processing and designing backaction-immune quantum metrology with nonlinearity.

quant-ph↗

arXiv2025

Language-Specific Layer Matters: Efficient Multilingual Enhancement for Large Vision-Language Models

Large vision-language models (LVLMs) have demonstrated exceptional capabilities in understanding visual information with human languages but also exhibit an imbalance in multilingual capabilities. In this work, we delve into the multilingual working pattern of LVLMs and identify a salient correlation between the multilingual understanding ability of LVLMs and language-specific neuron activations in shallow layers. Building on this insight, we introduce PLAST, a training recipe that achieves efficient multilingual enhancement for LVLMs by Precise LAnguage-Specific layers fine-Tuning. PLAST first identifies layers involved in multilingual understanding by monitoring language-specific neuron activations. These layers are then precisely fine-tuned with question-translation pairs to achieve multilingual alignment. Our empirical results on MM-Bench and MMMB demonstrate that PLAST effectively improves the multilingual capabilities of LVLMs and achieves significant efficiency with only 14% of the parameters tuned. Further analysis reveals that PLAST can be generalized to low-resource and complex visual reasoning tasks, facilitating the language-specific visual information engagement in shallow layers.

cs.CL↗

arXiv2025

Prototype: A Keyword Spotting-Based Intelligent Audio SoC for IoT

In this demo, we present a compact intelligent audio system-on-chip (SoC) integrated with a keyword spotting accelerator, enabling ultra-low latency, low-power, and low-cost voice interaction in Internet of Things (IoT) devices. Through algorithm-hardware co-design, the system's energy efficiency is maximized. We demonstrate the system's capabilities through a live FPGA-based prototype, showcasing stable performance and real-time voice interaction for edge intelligence applications.

cs.SD↗

arXiv2025

Action is All You Need: Dual-Flow Generative Ranking Network for Recommendation

Deep Learning Recommendation Models (DLRMs) often rely on extensive manual feature engineering to improve accuracy and user experience, which increases system complexity and limits scalability of model performance with respect to computational resources. Recently, Meta introduced a generative ranking paradigm based on HSTU block that enables end-to-end learning from raw user behavior sequences and demonstrates scaling law on large datasets that can be regarded as the state-of-the-art (SOTA). However, splitting user behaviors into interleaved item and action information significantly increases the input sequence length, which adversely affects both training and inference efficiency. To address this issue, we propose the Dual-Flow Generative Ranking Network (DFGR), that employs a dual-flow mechanism to optimize interaction modeling, ensuring efficient training and inference through end-to-end token processing. DFGR duplicates the original user behavior sequence into a real flow and a fake flow based on the authenticity of the action information, and then defines a novel interaction method between the real flow and the fake flow within the QKV module of the self-attention mechanism. This design reduces computational overhead and improves both training efficiency and inference performance compared to Meta's HSTU-based model. Experiments on both open-source and real industrial datasets show that DFGR outperforms DLRM, which serves as the industrial online baseline with extensive feature engineering, as well as Meta's HSTU and other common recommendation models such as DIN, DCN, DIEN, and DeepFM. Furthermore, we investigate optimal parameter allocation strategies under computational constraints, establishing DFGR as an efficient and effective next-generation generative ranking paradigm.

cs.IR↗

arXiv2025

Optimal Brain Connection: Towards Efficient Structural Pruning

Structural pruning has been widely studied for its effectiveness in compressing neural networks. However, existing methods often neglect the interconnections among parameters. To address this limitation, this paper proposes a structural pruning framework termed Optimal Brain Connection. First, we introduce the Jacobian Criterion, a first-order metric for evaluating the saliency of structural parameters. Unlike existing first-order methods that assess parameters in isolation, our criterion explicitly captures both intra-component interactions and inter-layer dependencies. Second, we propose the Equivalent Pruning mechanism, which utilizes autoencoders to retain the contributions of all original connection--including pruned ones--during fine-tuning. Experimental results demonstrate that the Jacobian Criterion outperforms several popular metrics in preserving model performance, while the Equivalent Pruning mechanism effectively mitigates performance degradation after fine-tuning. Code: https://github.com/ShaowuChen/Optimal_Brain_Connection

cs.CV↗

arXiv2025

EDGE-GRPO: Entropy-Driven GRPO with Guided Error Correction for Advantage Diversity

Large Language Models (LLMs) have made remarkable progress in enhancing step-by-step reasoning through reinforcement learning. However, the Group Relative Policy Optimization (GRPO) algorithm, which relies on sparse reward rules, often encounters the issue of identical rewards within groups, leading to the advantage collapse problem. Existing works typically address this challenge from two perspectives: enforcing model reflection to enhance response diversity, and introducing internal feedback to augment the training signal (advantage). In this work, we begin by analyzing the limitations of model reflection and investigating the policy entropy of responses at the fine-grained sample level. Based on our experimental findings, we propose the EDGE-GRPO algorithm, which adopts \textbf{E}ntropy-\textbf{D}riven Advantage and \textbf{G}uided \textbf{E}rror Correction to effectively mitigate the problem of advantage collapse. Extensive experiments on several main reasoning benchmarks demonstrate the effectiveness and superiority of our approach. It is available at https://github.com/ZhangXJ199/EDGE-GRPO.

cs.AI↗

arXiv2025

CoE-Ops: Collaboration of LLM-based Experts for AIOps Question-Answering

With the rapid evolution of artificial intelligence, AIOps has emerged as a prominent paradigm in DevOps. Lots of work has been proposed to improve the performance of different AIOps phases. However, constrained by domain-specific knowledge, a single model can only handle the operation requirement of a specific task,such as log parser,root cause analysis. Meanwhile, combining multiple models can achieve more efficient results, which have been proved in both previous ensemble learning and the recent LLM training domain. Inspired by these works,to address the similar challenges in AIOPS, this paper first proposes a collaboration-of-expert framework(CoE-Ops) incorporating a general-purpose large language model task classifier. A retrieval-augmented generation mechanism is introduced to improve the framework's capability in handling both Question-Answering tasks with high-level(Code,build,Test,etc.) and low-level(fault analysis,anomaly detection,etc.). Finally, the proposed method is implemented in the AIOps domain, and extensive experiments are conducted on the DevOps-EVAL dataset. Experimental results demonstrate that CoE-Ops achieves a 72% improvement in routing accuracy for high-level AIOps tasks compared to existing CoE methods, delivers up to 8% accuracy enhancement over single AIOps models in DevOps problem resolution, and outperforms larger-scale Mixture-of-Experts (MoE) models by up to 14% in accuracy.

cs.CL↗

arXiv2025

Subpixel correction of diffraction pattern shifts in ptychography via automatic differentiation

Ptychography, a coherent diffraction imaging technique, has become an indispensable tool in materials characterization, biological imaging, and nanostructure analysis due to its capability for high-resolution, lensless reconstruction of complex-valued images. In typical workflows, raw diffraction patterns are commonly cropped to isolate the valid central region before reconstruction. However, if the crop is misaligned from the diffraction pattern's zero-order, reconstruction may suffer from slower convergence, phase wrapping, and reduced image fidelity. These issues are further exacerbated in experimental configurations involving reflective geometries or broadband illumination, where incorrect cropping introduces systematic preprocessing errors that compromise the entire ptychographic inversion. To address this challenge, we present an approach based on automatic differentiation (AD), where the cropping shift is treated as an optimizable parameter within the reconstruction framework. By integrating shift correction into the backpropagation loop, our method simultaneously refines the object, probe, and shift positions without requiring manual tuning. Simulation results demonstrate that, even with initial offsets ranging up to 5 pixels, the proposed method achieves subpixel correction, with an average deviation below 0.5 pixels. Experiments in the extreme ultraviolet (EUV) regime further validate the method's robustness and effectiveness. This AD-based strategy enhances the automation and robustness of ptychographic reconstructions, and is adaptable to diverse experimental conditions.

physics.optics↗

arXiv2025

TinyLLaVA-Video: Towards Smaller LMMs for Video Understanding with Group Resampler

Video behavior recognition and scene understanding are fundamental tasks in multimodal intelligence, serving as critical building blocks for numerous real-world applications. Through large multimodal models (LMMs) have achieved remarkable progress in video understanding, most existing open-source models rely on over 7B parameters and require large-scale datasets for training, making them resource-intensive and inaccessible to many researchers. Furthermore, lightweight models face persistent challenges in effectively processing long visual sequences and temporal understanding. In this work, we introduce TinyLLaVA-Video, a lightweight yet powerful video understanding model with approximately 3.6B parameters. The cornerstone of our design is the video-level group resampler, a novel mechanism that significantly reduces and controls the number of visual tokens at the video level. Unlike traditional image-level resampler, our approach effectively mitigates redundancy while enhancing temporal comprehension, leading to improved performance on video-based tasks. In addition, TinyLLaVA-Video demonstrates exceptional efficiency, requiring only one day of training on 8 A100-40G GPUs. It surpasses several existing 7B-parameter models on multiple benchmarks. We believe this work provides a valuable foundation for future research on lightweight video understanding models. The code and weights is available at https://github.com/ZhangXJ199/TinyLLaVA-Video.

cs.CV↗

arXiv2025

TaxAgent: How Large Language Model Designs Fiscal Policy

Economic inequality is a global challenge, intensifying disparities in education, healthcare, and social stability. Traditional systems like the U.S. federal income tax reduce inequality but lack adaptability. Although models like the Saez Optimal Taxation adjust dynamically, they fail to address taxpayer heterogeneity and irrational behavior. This study introduces TaxAgent, a novel integration of large language models (LLMs) with agent-based modeling (ABM) to design adaptive tax policies. In our macroeconomic simulation, heterogeneous H-Agents (households) simulate real-world taxpayer behaviors while the TaxAgent (government) utilizes LLMs to iteratively optimize tax rates, balancing equity and productivity. Benchmarked against Saez Optimal Taxation, U.S. federal income taxes, and free markets, TaxAgent achieves superior equity-efficiency trade-offs. This research offers a novel taxation solution and a scalable, data-driven framework for fiscal policy evaluation.

cs.AI↗

arXiv2025

One for All: Update Parameterized Knowledge Across Multiple Models

Large language models (LLMs) encode vast world knowledge but struggle to stay up-to-date, often leading to errors and hallucinations. Knowledge editing offers an efficient alternative to retraining, enabling targeted modifications by updating specific model parameters. However, existing methods primarily focus on individual models, posing challenges in efficiently updating multiple models and adapting to new models. To address this, we propose OnceEdit, a novel ensemble-based approach that employs a plug-in model as the editing module, enabling stable knowledge updates across multiple models. Building on the model ensemble, OnceEdit introduces two key mechanisms to enhance its effectiveness. First, we introduce a dynamic weight mechanism through a \weight token for distinguishing between edit-related and non-edit-related instances, ensuring the appropriate utilization of knowledge from integrated models. Second, we incorporate an ensemble enhancement mechanism to mitigate the excessive reliance on the central model inherent in the model ensemble technique, making it more suitable for knowledge editing. Extensive experiments on diverse LLMs demonstrate that OnceEdit consistently outperforms existing methods while achieving superior editing efficiency. Further analysis confirms its adaptability and stability in multi-model editing scenarios. Our code will be available.

cs.CL↗