arXiv subjects

Zhihao Zhang

Publications and source records attributed to Zhihao Zhang.

At least 73 records · Page 4Linked to original sources

arXiv2025

Towards Efficient Generative Large Language Model Serving: A Survey from Algorithms to Systems

In the rapidly evolving landscape of artificial intelligence (AI), generative large language models (LLMs) stand at the forefront, revolutionizing how we interact with our data. However, the computational intensity and memory consumption of deploying these models present substantial challenges in terms of serving efficiency, particularly in scenarios demanding low latency and high throughput. This survey addresses the imperative need for efficient LLM serving methodologies from a machine learning system (MLSys) research perspective, standing at the crux of advanced AI innovations and practical system optimizations. We provide in-depth analysis, covering a spectrum of solutions, ranging from cutting-edge algorithmic modifications to groundbreaking changes in system designs. The survey aims to provide a comprehensive understanding of the current state and future directions in efficient LLM serving, offering valuable insights for researchers and practitioners in overcoming the barriers of effective LLM deployment, thereby reshaping the future of AI.

cs.LG↗

arXiv2025

DHEvo: Data-Algorithm Based Heuristic Evolution for Generalizable MILP Solving

Primal heuristics play a critical role in improving the efficiency of mixed integer programming (MILP) solvers. As large language models (LLMs) have demonstrated superior code generation abilities, recent MILP works are devoted to leveraging the evolutionary computation approaches with LLMs to generate effective primal heuristics. Although the generated heuristics have achieved better solving performance than the hand-crafted ones with little adaptability, the advantage of current LLM-based methods is limited to few MILP instances in one problem class, as they fail to capture the instance characteristics in the problem class (the MILP instances generated from the same mathematical model are defined as a problem class). Since MILP instances often differ significantly in structure and feature distribution, the neglect of their characteristics in the evolution process results in poor generalization within the same problem class. To overcome this challenge, we propose a data-algorithm co-evolution framework (DHEvo) that iteratively selects representative instances and evolves corresponding heuristics. With the initial instance distribution, we develop an LLM-based multi-agent system to generate data-code pairs simultaneously. These data-code pairs are iteratively refined based on their fitness scores, leading to the identification of the most effective heuristic over the entire problem class. Extensive experiments across diverse MILP benchmarks demonstrate that our approach significantly outperforms both human-designed heuristics and existing LLM-based methods.

cs.NE↗

arXiv2025

Test mass charge management in the detection of gravitational waves in space based on UV micro-LED

As an alternative to the ultraviolet light emitting diode(UV LED), the feasibility of utilizing UV micro-LED in the charge management in the detection of gravitational waves in space is experimentally studied. Compared with UV LED, micro-LED is more compact in size, has better current spreading, faster response time and longer operating life. Performance characteristics of micro-LEDs were measured, with peak wavelength of 254 nm, 262 nm, 274 nm, and 282 nm for each respective micro-LED, and the photoelectric effect was demonstrated. The effectiveness of micro-LED based charge management experiments were demonstrated using above micro-LEDs mounted on a cubical test mass, and different discharge rates were achieved by varying the drive current and duty cycle using pulse width modulation(PWM). Laboratory data was also shown to demonstrate the space qualification of the micro-LED device, the key electrical and optical characteristics of the micro-LEDs showed less than 5% variation. The results of the qualification bring the micro-LED device Technology Readiness Level(TRL) to TRL-5. TRL-6 will be reached provided additional radiation and thermal tests are conducted and in a position ready to be flown and further tested in space.

physics.ins-det↗

arXiv2025

Tile Classification Based Viewport Prediction with Multi-modal Fusion Transformer

Viewport prediction is a crucial aspect of tile-based 360 video streaming system. However, existing trajectory based methods lack of robustness, also oversimplify the process of information construction and fusion between different modality inputs, leading to the error accumulation problem. In this paper, we propose a tile classification based viewport prediction method with Multi-modal Fusion Transformer, namely MFTR. Specifically, MFTR utilizes transformer-based networks to extract the long-range dependencies within each modality, then mine intra- and inter-modality relations to capture the combined impact of user historical inputs and video contents on future viewport selection. In addition, MFTR categorizes future tiles into two categories: user interested or not, and selects future viewport as the region that contains most user interested tiles. Comparing with predicting head trajectories, choosing future viewport based on tile's binary classification results exhibits better robustness and interpretability. To evaluate our proposed MFTR, we conduct extensive experiments on two widely used PVS-HM and Xu-Gaze dataset. MFTR shows superior performance over state-of-the-art methods in terms of average prediction accuracy and overlap ratio, also presents competitive computation efficiency.

cs.CV↗

arXiv2025

PFDial: A Structured Dialogue Instruction Fine-tuning Method Based on UML Flowcharts

Process-driven dialogue systems, which operate under strict predefined process constraints, are essential in customer service and equipment maintenance scenarios. Although Large Language Models (LLMs) have shown remarkable progress in dialogue and reasoning, they still struggle to solve these strictly constrained dialogue tasks. To address this challenge, we construct Process Flow Dialogue (PFDial) dataset, which contains 12,705 high-quality Chinese dialogue instructions derived from 440 flowcharts containing 5,055 process nodes. Based on PlantUML specification, each UML flowchart is converted into atomic dialogue units i.e., structured five-tuples. Experimental results demonstrate that a 7B model trained with merely 800 samples, and a 0.5B model trained on total data both can surpass 90% accuracy. Additionally, the 8B model can surpass GPT-4o up to 43.88% with an average of 11.00%. We further evaluate models' performance on challenging backward transitions in process flows and conduct an in-depth analysis of various dataset formats to reveal their impact on model performance in handling decision and sequential branches. The data is released in https://github.com/KongLongGeFDU/PFDial.

cs.CL↗

arXiv2025

Rogue waves collision under incident momentum modulation in two-component Bose-Einstein condensates

The collision dynamics of two first-order rogue waves (RWs) with opposite incident momentum in two-component Bose-Einstein condensates (BECs) is studied by solving the two-component one-dimensional Gross-Pitaevskii (GP) equation. It is demonstrated that the introduction of appropriate incident momentum successfully promotes the generation of second-order RWs in the case of relatively weaker interspecies interactions compared to intraspecific interactions. The range of incident momentum that can facilitate the generation of second-order RWs under different interspecies interaction strengths is determined, and machine learning is employed to find and analyze relationships among the interspecies interaction, the incident momentum, and the offset that can lead to the generation of second-order RWs. It shows that any two parameters above exhibit a positive or negative correlation when the third parameter is fixed. These findings provide additional possibilities for generating and controlling high-order RWs.

cond-mat.quant-gas↗

arXiv2025

Reinforcement Learning with Data Bootstrapping for Dynamic Subgoal Pursuit in Humanoid Robot Navigation

Safe and real-time navigation is fundamental for humanoid robot applications. However, existing bipedal robot navigation frameworks often struggle to balance computational efficiency with the precision required for stable locomotion. We propose a novel hierarchical framework that continuously generates dynamic subgoals to guide the robot through cluttered environments. Our method comprises a high-level reinforcement learning (RL) planner for subgoal selection in a robot-centric coordinate system and a low-level Model Predictive Control (MPC) based planner which produces robust walking gaits to reach these subgoals. To expedite and stabilize the training process, we incorporate a data bootstrapping technique that leverages a model-based navigation approach to generate a diverse, informative dataset. We validate our method in simulation using the Agility Robotics Digit humanoid across multiple scenarios with random obstacles. Results show that our framework significantly improves navigation success rates and adaptability compared to both the original model-based method and other learning-based methods.

cs.RO↗

arXiv2025

AdaServe: Accelerating Multi-SLO LLM Serving with SLO-Customized Speculative Decoding

Modern large language model (LLM) applications exhibit diverse service-level objectives (SLOs), from low-latency requirements in interactive coding assistants to more relaxed constraints in data wrangling tasks. Existing LLM serving systems, which rely on uniform batching and scheduling strategies, often fail to meet these heterogeneous SLOs concurrently. We present AdaServe, the first LLM serving system designed to support efficient multi-SLO serving through SLO-customized speculative decoding. AdaServe formulates multi-SLO serving as a constrained optimization problem and introduces a hardware-aware algorithm that constructs a speculation tree tailored to each request's latency target. It features a speculate-select-verify pipeline that enables fine-grained control over decoding speed while maximizing system throughput. AdaServe further adapts to workload variation by dynamically adjusting speculation parameters. Evaluations across diverse workloads show that AdaServe reduces SLO violations by up to 4.3$\times$ and improves goodput by up to 1.9$\times$ compared to the best performing baselines, highlighting its effectiveness in multi-SLO serving.

cs.CL↗

arXiv2025

SpecReason: Fast and Accurate Inference-Time Compute via Speculative Reasoning

Recent advances in inference-time compute have significantly improved performance on complex tasks by generating long chains of thought (CoTs) using Large Reasoning Models (LRMs). However, this improved accuracy comes at the cost of high inference latency due to the length of generated reasoning sequences and the autoregressive nature of decoding. Our key insight in tackling these overheads is that LRM inference, and the reasoning that it embeds, is highly tolerant of approximations: complex tasks are typically broken down into simpler steps, each of which brings utility based on the semantic insight it provides for downstream steps rather than the exact tokens it generates. Accordingly, we introduce SpecReason, a system that automatically accelerates LRM inference by using a lightweight model to (speculatively) carry out simpler intermediate reasoning steps and reserving the costly base model only to assess (and potentially correct) the speculated outputs. Importantly, SpecReason's focus on exploiting the semantic flexibility of thinking tokens in preserving final-answer accuracy is complementary to prior speculation techniques, most notably speculative decoding, which demands token-level equivalence at each step. Across a variety of reasoning benchmarks, SpecReason achieves $1.4-3.0\times$ speedup over vanilla LRM inference while improving accuracy by $0.4-9.0\%$. Compared to speculative decoding without SpecReason, their combination yields an additional $8.8-58.0\%$ latency reduction. We open-source SpecReason at https://github.com/ruipeterpan/specreason.

cs.LG↗

arXiv2025

Wiggly dilaton: a landscape of spontaneously broken scale invariance

The dilaton emerges as a pseudo-Nambu-Goldstone boson (pNGB) associated with the spontaneous breaking of scale invariance in a nearly conformal field theory (CFT). We show the existence of a wiggly dilaton potential that contains multiple vacuum solutions in a five-dimensional (5D) holographic formulation. The wiggly feature originates from boundary potentials of a 5D axion-like scalar field, whose naturally small bulk mass parameter corresponds to a marginally-relevant deformation of the dual CFT. Depending on the energy density of a boundary 3-brane, our model can provide a relaxion potential or generate a light dilaton. However, an extremely light dilaton requires fine-tuning.

hep-th↗

arXiv2025

SEB-Naver: A SE(2)-based Local Navigation Framework for Car-like Robots on Uneven Terrain

Autonomous navigation of car-like robots on uneven terrain poses unique challenges compared to flat terrain, particularly in traversability assessment and terrain-associated kinematic modelling for motion planning. This paper introduces SEB-Naver, a novel SE(2)-based local navigation framework designed to overcome these challenges. First, we propose an efficient traversability assessment method for SE(2) grids, leveraging GPU parallel computing to enable real-time updates and maintenance of local maps. Second, inspired by differential flatness, we present an optimization-based trajectory planning method that integrates terrain-associated kinematic models, significantly improving both planning efficiency and trajectory quality. Finally, we unify these components into SEB-Naver, achieving real-time terrain assessment and trajectory optimization. Extensive simulations and real-world experiments demonstrate the effectiveness and efficiency of our approach. The code is at https://github.com/ZJU-FAST-Lab/seb_naver.

cs.RO↗

arXiv2025

EliteKV: Scalable KV Cache Compression via RoPE Frequency Selection and Joint Low-Rank Projection

Rotary Position Embedding (RoPE) enables each attention head to capture multi-frequency information along the sequence dimension and is widely applied in foundation models. However, the nonlinearity introduced by RoPE complicates optimization of the key state in the Key-Value (KV) cache for RoPE-based attention. Existing KV cache compression methods typically store key state before rotation and apply the transformation during decoding, introducing additional computational overhead. This paper introduces EliteKV, a flexible modification framework for RoPE-based models supporting variable KV cache compression ratios. EliteKV first identifies the intrinsic frequency preference of each head using RoPElite, selectively restoring linearity to certain dimensions of key within attention computation. Building on this, joint low-rank compression of key and value enables partial cache sharing. Experimental results show that with minimal uptraining on only $0.6\%$ of the original training data, RoPE-based models achieve a $75\%$ reduction in KV cache size while preserving performance within a negligible margin. Furthermore, EliteKV consistently performs well across models of different scales within the same family.

cs.LG↗

arXiv2025

RU-AI: A Large Multimodal Dataset for Machine-Generated Content Detection

The recent generative AI models' capability of creating realistic and human-like content is significantly transforming the ways in which people communicate, create and work. The machine-generated content is a double-edged sword. On one hand, it can benefit the society when used appropriately. On the other hand, it may mislead people, posing threats to the society, especially when mixed together with natural content created by humans. Hence, there is an urgent need to develop effective methods to detect machine-generated content. However, the lack of aligned multimodal datasets inhibited the development of such methods, particularly in triple-modality settings (e.g., text, image, and voice). In this paper, we introduce RU-AI, a new large-scale multimodal dataset for robust and effective detection of machine-generated content in text, image and voice. Our dataset is constructed on the basis of three large publicly available datasets: Flickr8K, COCO and Places205, by adding their corresponding AI duplicates, resulting in a total of 1,475,370 instances. In addition, we created an additional noise variant of the dataset for testing the robustness of detection models. We conducted extensive experiments with the current SOTA detection methods on our dataset. The results reveal that existing models still struggle to achieve accurate and robust detection on our dataset. We hope that this new data set can promote research in the field of machine-generated content detection, fostering the responsible use of generative AI. The source code and datasets are available at https://github.com/ZhihaoZhang97/RU-AI.

cs.CV↗

arXiv2025

LLMs can Realize Combinatorial Creativity: Generating Creative Ideas via LLMs for Scientific Research

Scientific idea generation has been extensively studied in creativity theory and computational creativity research, providing valuable frameworks for understanding and implementing creative processes. However, recent work using Large Language Models (LLMs) for research idea generation often overlooks these theoretical foundations. We present a framework that explicitly implements combinatorial creativity theory using LLMs, featuring a generalization-level retrieval system for cross-domain knowledge discovery and a structured combinatorial process for idea generation. The retrieval system maps concepts across different abstraction levels to enable meaningful connections between disparate domains, while the combinatorial process systematically analyzes and recombines components to generate novel solutions. Experiments on the OAG-Bench dataset demonstrate our framework's effectiveness, consistently outperforming baseline approaches in generating ideas that align with real research developments (improving similarity scores by 7\%-10\% across multiple metrics). Our results provide strong evidence that LLMs can effectively realize combinatorial creativity when guided by appropriate theoretical frameworks, contributing both to practical advancement of AI-assisted research and theoretical understanding of machine creativity.

cs.AI↗

arXiv2025

Spin Transport Hydrodynamics of Polarized Deuterium-Tritium Fusion Plasma

The spin transport equations for polarized deuterium-tritium (DT) fusion plasma are derived with the density matrix formulation, which are used to investigate the hydrodynamics of polarized DT-gas-filled targets during indirectly driven inertial confinement fusion implosions. The depolarization of DT ions by strong self-generated magnetic fields can be captured by the spin transport equation. The neutron yield, angular distribution and neutron beam polarization are obtained from three-dimensional spin transport hydrodynamics simulations of the target implosions. The simulation results indicate that an optimized spin alignment of the polarized target can reduce the depolarization of DT ions and the neutron beams induced by polar mode asymmetries in indirectly driven implosions.

physics.plasm-ph↗

arXiv2025

Extensive Exploration in Complex Traffic Scenarios using Hierarchical Reinforcement Learning

Developing an automated driving system capable of navigating complex traffic environments remains a formidable challenge. Unlike rule-based or supervised learning-based methods, Deep Reinforcement Learning (DRL) based controllers eliminate the need for domain-specific knowledge and datasets, thus providing adaptability to various scenarios. Nonetheless, a common limitation of existing studies on DRL-based controllers is their focus on driving scenarios with simple traffic patterns, which hinders their capability to effectively handle complex driving environments with delayed, long-term rewards, thus compromising the generalizability of their findings. In response to these limitations, our research introduces a pioneering hierarchical framework that efficiently decomposes intricate decision-making problems into manageable and interpretable subtasks. We adopt a two step training process that trains the high-level controller and low-level controller separately. The high-level controller exhibits an enhanced exploration potential with long-term delayed rewards, and the low-level controller provides longitudinal and lateral control ability using short-term instantaneous rewards. Through simulation experiments, we demonstrate the superiority of our hierarchical controller in managing complex highway driving situations.

cs.LG↗

arXiv2025

Communication Bounds for the Distributed Experts Problem

In this work, we study the experts problem in the distributed setting where an expert's cost needs to be aggregated across multiple servers. Our study considers various communication models such as the message-passing model and the broadcast model, along with multiple aggregation functions, such as summing and taking the $\ell_p$ norm of an expert's cost across servers. We propose the first communication-efficient protocols that achieve near-optimal regret in these settings, even against a strong adversary who can choose the inputs adaptively. Additionally, we give a conditional lower bound showing that the communication of our protocols is nearly optimal. Finally, we implement our protocols and demonstrate empirical savings on the HPO-B benchmarks.

cs.LG↗

arXiv2024

A Claim Decomposition Benchmark for Long-form Answer Verification

The advancement of LLMs has significantly boosted the performance of complex long-form question answering tasks. However, one prominent issue of LLMs is the generated "hallucination" responses that are not factual. Consequently, attribution for each claim in responses becomes a common solution to improve the factuality and verifiability. Existing researches mainly focus on how to provide accurate citations for the response, which largely overlook the importance of identifying the claims or statements for each response. To bridge this gap, we introduce a new claim decomposition benchmark, which requires building system that can identify atomic and checkworthy claims for LLM responses. Specifically, we present the Chinese Atomic Claim Decomposition Dataset (CACDD), which builds on the WebCPM dataset with additional expert annotations to ensure high data quality. The CACDD encompasses a collection of 500 human-annotated question-answer pairs, including a total of 4956 atomic claims. We further propose a new pipeline for human annotation and describe the challenges of this task. In addition, we provide experiment results on zero-shot, few-shot and fine-tuned LLMs as baselines. The results show that the claim decomposition is highly challenging and requires further explorations. All code and data are publicly available at \url{https://github.com/FBzzh/CACDD}.

cs.CL↗