arXiv subjects

Yifu Ding

Publications and source records attributed to Yifu Ding.

At least 19 recordsLinked to original sources

arXiv2026

Refined Thompson Learning for Adaptive Bandits: Power-Efficient Flexibility Scheduling Across Data Centers

The rapid growth of large-scale AI workloads in data centers has placed increasing pressure on power grids in recent years. Since power systems must continuously balance supply and demand, there is growing interests in leveraging data-center workload flexibility as a grid service. We propose a contextual restless multi-armed bandit (CRMAB) framework in which a grid operator requests load reductions without observing internal job-scheduling decisions. Under index-ability guarantee, each data center or physical machine is modeled as a Markov decision process (MDP) over a cyclic virtual-machine (VM) job queue, with unknown rewards and transition dynamics learned online using Thompson sampling and Whittle-index policies. To improve learning under sparse and noisy observations, the framework augments an adaptive Thompson--Whittle (TW) policy with domain-informed transition priors and gated prior mixing. In baseline experiments, the best adaptive refined variant achieves 91.4\% of the oracle reward after 100 rounds and 96.8\% after 1,000 rounds. Across a 16-setting stress test spanning different state-space sizes and levels of contextual noise, the best refined variant consistently outperforms the original TW policy with high confidence while remaining competitive with EXP4. A graph-based prior further incorporates data-center hardware constraints, including computing-resource limits. Overall, the results demonstrate the economic potential of data-center flexibility as a grid service and highlight the importance of high-quality, open-source AI workload traces for developing and evaluating such services.

cs.CE↗

arXiv2026

Multi-Resolution Flow Matching: Training-Free Diffusion Acceleration via Staged Sampling

Hardware-agnostic strategies for accelerating text-to-image diffusion, such as timestep distillation and feature caching, can reduce inference time without custom kernels or system-level optimization. Among them, multi-resolution generation strategies have recently received broad attention, attaining more than 5x speedup without any training. However, the design of performing upsampling in the latent space, together with the selective modification of partial regions, causes these methods to exhibit noticeable blurring or artifacts. To this end, we propose MrFlow, a training-free multi-resolution acceleration strategy for pretrained flow-matching models built upon a staged low-to-high-resolution pipeline. MrFlow first rapidly generates the main structure at low resolution, then performs super-resolution in the pixel space using a lightweight pretrained GAN-based model, subsequently injects low-strength noise to enable high-frequency resampling, and finally refines the details at high resolution. Quantitative and qualitative results on FLUX.1-dev and Qwen-Image show that MrFlow exploits the quadratic token reduction and reduced step requirement of low-resolution sampling to achieve 10x end-to-end acceleration while keeping OneIG within a 1% gap relative to that before acceleration, significantly surpassing other training-free acceleration strategies, and requiring no training or runtime dynamic identification whatsoever. MrFlow can further be directly combined orthogonally with pre-trained timestep distillation strategies, achieving even higher generation acceleration of up to 25x.

cs.CV↗

arXiv2026

Attribution-Guided and Coverage-Maximized Pruning for Structural MoE Compression

Mixture-of-Experts (MoE) models scale compute efficiently, yet remain expensive to deploy due to their substantial memory footprint and inference overhead. Prior compression methods mainly operate at the expert level, either removing entire experts or ranking experts by coarse-grained importance scores. However, such expert-wise decisions are often too coarse to capture fine-grained redundancy, leading to misallocated pruning budgets and limited compression. To address this problem, we observe that information within MoE experts is highly concentrated in a small subset of channels, leaving substantial redundancy even in experts deemed important. Based on this observation, we propose a structural pruning framework tailored for MoE models. Our method reformulates prune-ratio allocation as a channel-score coverage maximization problem and solves it efficiently using an attribution-based approximation. Experiments on DeepSeek and Qwen MoE models show that our method preserves model accuracy under 50% or 25% structured pruning when combined with 4-bit quantization. On Qwen3-30B-A3B, our approach reduces memory footprint by 5.27$\times$ and consistently outperforms state-of-the-art baselines across diverse benchmarks.

cs.LG↗

arXiv2026

Robust Restless Multi-Armed Bandit for Data Center Flexibility Services Through Virtual Machine Scheduling

Energy demands from data centers have surged and stressed the grid in recent years. Electric grids require balancing supply and demand every second, motivating demand response (reduction) from large loads, including data centers. This can be achieved by rescheduling jobs on a physical machine. Its real-time implementation is uncertain due to fluctuating resource utilization, and rescheduling incurs quality-of-service (QoS) losses that providers are unwilling to disclose. We propose a restless multi-armed bandit (RMAB) framework, in which the grid operator requests load reductions without access to detailed job-rescheduling procedures. Using open-source virtual machine (VM) datasets, we model job arrivals and rescheduling at each data center as a restless arm in a Markov decision process (MDP) and derive Whittle-index-based policies using the learned transition function via Thompson sampling. To overcome the weakness of an increasingly long learning process due to an enlarged state space, we use a mixed strategy that includes a global upper confidence bound (UCB) and encodes trust indices to enhance robustness and accelerate learning. Results show that the proposed mixed-strategy algorithm remains robust across varying state-space sizes and consistently outperforms the pure Thompson-Whittle (TW) algorithm, especially when contextual information is noisy. It also demonstrates superior performance compared to the state-of-the-art EXP4 framework. We provided open-source code to ensure reproducibility.

cs.CE↗

arXiv2026

Diagonal-Tiled Mixed-Precision Attention for Efficient Low-Bit MXFP Inference

Transformer-based large language models (LLMs) have demonstrated remarkable performance across a wide range of real-world tasks, but their inference cost remains prohibitively high due to the quadratic complexity of attention and the memory bandwidth limitations of high-precision operations. In this work, we present a low-bit mixed-precision attention kernel using the microscaling floating-point (MXFP) data format, utilizing the computing capability on next-generation GPU architectures. Our Diagonal-Tiled Mixed-Precision Attention (DMA) incorporates two kinds of low-bit computation at the tiling-level, and is a delicate fused kernel implemented using Triton, exploiting hardware-level parallelism and memory efficiency to enable fast and efficient inference without compromising model performance. Extensive empirical evaluations on NVIDIA B200 GPUs show that our kernel maintains generation quality with negligible degradation, and meanwhile achieves significant speedup by kernel fusion. We release our code at https://github.com/yifu-ding/MP-Sparse-Attn.

cs.LG↗

arXiv2026

BWTA: Accurate and Efficient Binarized Transformer by Algorithm-Hardware Co-design

Ultra low-bit quantization brings substantial efficiency for Transformer-based models, but the accuracy degradation and limited GPU support hinder its wide usage. In this paper, we analyze zero-point distortion in binarization and propose a Binary Weights & Ternary Activations (BWTA) quantization scheme, which projects tiny values to zero and preserves the accuracy of extremely low-bit models. For training, we propose Smooth Multi-Stage Quantization, combining a Levelwise Degradation Strategy and a Magnitude-Alignment Projection Factor to enable stable and fast convergence. For inference, we develop a BWTA MatMul CUDA kernel with instruction-level parallel bit-packing and comprehensive binary/ternary MatMul implementations for both linear and attention operators, allowing seamless integration across Transformer architectures. Experiments show that BWTA approaches full-precision performance for BERT, with an average 3.5% drop on GLUE and less than 2% drop on five tasks, and achieves comparable perplexity and accuracy for LLMs. In efficiency, it delivers 16 to 24 times kernel-level speedup over FP16 on NVIDIA GPUs, and 216 to 330 tokens/s end-to-end prefill speedup with lower memory footprint on LLMs. As an algorithm-hardware co-design, BWTA demonstrates practical, low-latency ultra-low-bit inference without sacrificing model quality.

cs.LG↗

arXiv2026

SPA-Cache: Singular Proxies for Adaptive Caching in Diffusion Language Models

While Diffusion Language Models (DLMs) offer a flexible, arbitrary-order alternative to the autoregressive paradigm, their non-causal nature precludes standard KV caching, forcing costly hidden state recomputation at every decoding step. Existing DLM caching approaches reduce this cost by selective hidden state updates; however, they are still limited by (i) costly token-wise update identification heuristics and (ii) rigid, uniform budget allocation that fails to account for heterogeneous hidden state dynamics. To address these challenges, we present SPA-Cache that jointly optimizes update identification and budget allocation in DLM cache. First, we derive a low-dimensional singular proxy that enables the identification of update-critical tokens in a low-dimensional subspace, substantially reducing the overhead of update identification. Second, we introduce an adaptive strategy that allocates fewer updates to stable layers without degrading generation quality. Together, these contributions significantly improve the efficiency of DLMs, yielding up to an $8\times$ throughput improvement over vanilla decoding and a $2$--$4\times$ speedup over existing caching baselines.

cs.LG↗

arXiv2025

MoDES: Accelerating Mixture-of-Experts Multimodal Large Language Models via Dynamic Expert Skipping

Mixture-of-Experts (MoE) Multimodal large language models (MLLMs) excel at vision-language tasks, but they suffer from high computational inefficiency. To reduce inference overhead, expert skipping methods have been proposed to deactivate redundant experts based on the current input tokens. However, we find that applying these methods-originally designed for unimodal large language models (LLMs)-to MLLMs results in considerable performance degradation. This is primarily because such methods fail to account for the heterogeneous contributions of experts across MoE layers and modality-specific behaviors of tokens within these layers. Motivated by these findings, we propose MoDES, the first training-free framework that adaptively skips experts to enable efficient and accurate MoE MLLM inference. It incorporates a globally-modulated local gating (GMLG) mechanism that integrates global layer-wise importance into local routing probabilities to accurately estimate per-token expert importance. A dual-modality thresholding (DMT) method is then applied, which processes tokens from each modality separately, to derive the skipping schedule. To set the optimal thresholds, we introduce a frontier search algorithm that exploits monotonicity properties, cutting convergence time from several days to a few hours. Extensive experiments for 3 model series across 13 benchmarks demonstrate that MoDES far outperforms previous approaches. For instance, when skipping 88% experts for Qwen3-VL-MoE-30B-A3B-Instruct, the performance boost is up to 10.67% (97.33% vs. 86.66%). Furthermore, MoDES significantly enhances inference speed, improving the prefilling time by 2.16$\times$ and the decoding time by 1.26$\times$. Our code is available at https://github.com/ModelTC/MoDES.

cs.CV↗

arXiv2025

Decision-dependent Robust Charging Infrastructure Planning for Light-duty Truck Electrification at Industrial Sites

Many industrial sites and digital logistics platforms rely on diesel-powered light-duty trucks to transport workers and small-scale facilities, which results in a significant amount of greenhouse gas emissions (GHGs). To address this, we develop a robust model for planning charging infrastructure to electrify light-duty trucks at industrial sites. The model is formulated as a mixed-integer linear program (MILP) that optimizes the charging infrastructure selection (across multiple charger types and locations) and determines charging schedules for each truck based on the selected infrastructure. Given the strict stop times and schedules at industrial sites, we introduce a scheduling-with-abandonment problem in which trucks forgo charging if their waiting time exceeds a maximum threshold. We further incorporate the impacts of overnight charging and range anxiety on drivers' waiting and abandonment behaviors. To model stochastic, heterogeneous parking durations, we classified trucks using machine learning (ML) methods based on contextual and time-location features. We then constructed decision-dependent, feature-driven robust uncertainty sets in which parking-time variability varies flexibly with drivers' charging choices. These feature-driven sets are applied to two robust optimization formulations with decision-dependent uncertainty (RO-DDU), resulting in distinct outcomes and managerial implications. We conduct a case study at an open-pit mining site to plan charger installations across eight charging zones, serving approximately 200 trucks. By decomposing the problem into a short rolling horizon or using a heuristic approach for the full-year or representative-day dataset, the model achieves an optimality gap of less than 0.1\% under diverse uncertainty scenarios.

eess.SY↗

arXiv2025

VORTA: Efficient Video Diffusion via Routing Sparse Attention

Video diffusion transformers have achieved remarkable progress in high-quality video generation, but remain computationally expensive due to the quadratic complexity of attention over high-dimensional video sequences. Recent acceleration methods enhance the efficiency by exploiting the local sparsity of attention scores; yet they often struggle with accelerating the long-range computation. To address this problem, we propose VORTA, an acceleration framework with two novel components: 1) a sparse attention mechanism that efficiently captures long-range dependencies, and 2) a routing strategy that adaptively replaces full 3D attention with specialized sparse attention variants. VORTA achieves an end-to-end speedup $1.76\times$ without loss of quality on VBench. Furthermore, it can seamlessly integrate with various other acceleration methods, such as model caching and step distillation, reaching up to speedup $14.41\times$ with negligible performance degradation. VORTA demonstrates its efficiency and enhances the practicality of video diffusion transformers in real-world settings. Codes and weights are available at https://github.com/wenhao728/VORTA.

cs.CV↗

arXiv2025

QVGen: Pushing the Limit of Quantized Video Generative Models

Video diffusion models (DMs) have enabled high-quality video synthesis. Yet, their substantial computational and memory demands pose serious challenges to real-world deployment, even on high-end GPUs. As a commonly adopted solution, quantization has proven notable success in reducing cost for image DMs, while its direct application to video DMs remains ineffective. In this paper, we present QVGen, a novel quantization-aware training (QAT) framework tailored for high-performance and inference-efficient video DMs under extremely low-bit quantization (e.g., 4-bit or below). We begin with a theoretical analysis demonstrating that reducing the gradient norm is essential to facilitate convergence for QAT. To this end, we introduce auxiliary modules ($\Phi$) to mitigate large quantization errors, leading to significantly enhanced convergence. To eliminate the inference overhead of $\Phi$, we propose a rank-decay strategy that progressively eliminates $\Phi$. Specifically, we repeatedly employ singular value decomposition (SVD) and a proposed rank-based regularization $\mathbf{\gamma}$ to identify and decay low-contributing components. This strategy retains performance while zeroing out additional inference overhead. Extensive experiments across $4$ state-of-the-art (SOTA) video DMs, with parameter sizes ranging from $1.3\text{B}\sim14\text{B}$, show that QVGen is the first to reach full-precision comparable quality under 4-bit settings. Moreover, it significantly outperforms existing methods. For instance, our 3-bit CogVideoX-2B achieves improvements of $+25.28$ in Dynamic Degree and $+8.43$ in Scene Consistency on VBench. Code and models are available at https://github.com/ModelTC/QVGen.

cs.CV↗

arXiv2025

Dynamic Parallel Tree Search for Efficient LLM Reasoning

Tree of Thoughts (ToT) enhances Large Language Model (LLM) reasoning by structuring problem-solving as a spanning tree. However, recent methods focus on search accuracy while overlooking computational efficiency. The challenges of accelerating the ToT lie in the frequent switching of reasoning focus, and the redundant exploration of suboptimal solutions. To alleviate this dilemma, we propose Dynamic Parallel Tree Search (DPTS), a novel parallelism framework that aims to dynamically optimize the reasoning path in inference. It includes the Parallelism Streamline in the generation phase to build up a flexible and adaptive parallelism with arbitrary paths by fine-grained cache management and alignment. Meanwhile, the Search and Transition Mechanism filters potential candidates to dynamically maintain the reasoning focus on more possible solutions and have less redundancy. Experiments on Qwen-2.5 and Llama-3 with Math500 and GSM8K datasets show that DPTS significantly improves efficiency by 2-4x on average while maintaining or even surpassing existing reasoning algorithms in accuracy, making ToT-based reasoning more scalable and computationally efficient.

cs.AI↗

arXiv2024

LLMCBench: Benchmarking Large Language Model Compression for Efficient Deployment

Although large language models (LLMs) have demonstrated their strong intelligence ability, the high demand for computation and storage hinders their practical application. To this end, many model compression techniques are proposed to increase the efficiency of LLMs. However, current researches only validate their methods on limited models, datasets, metrics, etc, and still lack a comprehensive evaluation under more general scenarios. So it is still a question of which model compression approach we should use under a specific case. To mitigate this gap, we present the Large Language Model Compression Benchmark (LLMCBench), a rigorously designed benchmark with an in-depth analysis for LLM compression algorithms. We first analyze the actual model production requirements and carefully design evaluation tracks and metrics. Then, we conduct extensive experiments and comparison using multiple mainstream LLM compression approaches. Finally, we perform an in-depth analysis based on the evaluation and provide useful insight for LLM compression design. We hope our LLMCBench can contribute insightful suggestions for LLM compression algorithm design and serve as a foundation for future research. Our code is available at https://github.com/AboveParadise/LLMCBench.

cs.CL↗

arXiv2024

A Survey of Low-bit Large Language Models: Basics, Systems, and Algorithms

Large language models (LLMs) have achieved remarkable advancements in natural language processing, showcasing exceptional performance across various tasks. However, the expensive memory and computational requirements present significant challenges for their practical deployment. Low-bit quantization has emerged as a critical approach to mitigate these challenges by reducing the bit-width of model parameters, activations, and gradients, thus decreasing memory usage and computational demands. This paper presents a comprehensive survey of low-bit quantization methods tailored for LLMs, covering the fundamental principles, system implementations, and algorithmic strategies. An overview of basic concepts and new data formats specific to low-bit LLMs is first introduced, followed by a review of frameworks and systems that facilitate low-bit LLMs across various hardware platforms. Then, we categorize and analyze techniques and toolkits for efficient low-bit training and inference of LLMs. Finally, we conclude with a discussion of future trends and potential advancements of low-bit LLMs. Our systematic overview from basic, system, and algorithm perspectives can offer valuable insights and guidelines for future works to enhance the efficiency and applicability of LLMs through low-bit quantization.

cs.AI↗

arXiv2024

The role of coal plant retrofitting strategies in developing India's net-zero power system: a data-driven sub-national analysis

India set two Nationally Determined Contribution targets to achieve the net zero carbon emission goal by 2070, which requires deep decarbonization of India's power generation sector. Yet, coal power generation contributes to more than 60\% of its total power generation, and policies still permit further coal fleet expansion and lifetime extensions. In this paper, we investigate the role of retrofitting India's coal plants for carbon capture and storage and biomass co-firing in developing the net-zero power system. We model the power generation and transmission network expansions across 30 Indian states in four representative grid evolution scenarios under progressively tighter carbon emission caps, taking into account sub-national coal price variation and thermal efficiency of individual coal plants. We find that coal plant retrofitting could happen by 2035 when an annual carbon cap for the power generation sector is less than 1,000 million tons CO$_2$. This reduces the unabated coal plant capacity, electricity generation, and carbon abatement costs. Exploiting renewable energy potentials solely, such as wind resources, could reduce carbon abatement costs significantly but will result in low coal plant utilization and uneven renewable generation deployment between Southern and Central regions concerning energy justice.

physics.soc-ph↗

arXiv2024

A Dataset of the Operating Station Heat Rate for 806 Indian Coal Plant Units using Machine Learning

India aims to achieve net-zero emissions by 2070 and has set an ambitious target of 500 GW of renewable power generation capacity by 2030. Coal plants currently contribute to more than 60\% of India's electricity generation in 2022. Upgrading and decarbonizing high-emission coal plants became a pressing energy issue. A key technical parameter for coal plants is the operating station heat rate (SHR), which represents the thermal efficiency of a coal plant. Yet, the operating SHR of Indian coal plants varies and is not comprehensively documented. This study extends from several existing databases and creates an SHR dataset for 806 Indian coal plant units using machine learning (ML), presenting the most comprehensive coverage to date. Additionally, it incorporates environmental factors such as water stress risk and coal prices as prediction features to improve accuracy. This dataset, easily downloadable from our visualization platform, could inform energy and environmental policies for India's coal power generation as the country transitions towards its renewable energy targets.

cs.CY↗

arXiv2024

PTQ4SAM: Post-Training Quantization for Segment Anything

Segment Anything Model (SAM) has achieved impressive performance in many computer vision tasks. However, as a large-scale model, the immense memory and computation costs hinder its practical deployment. In this paper, we propose a post-training quantization (PTQ) framework for Segment Anything Model, namely PTQ4SAM. First, we investigate the inherent bottleneck of SAM quantization attributed to the bimodal distribution in post-Key-Linear activations. We analyze its characteristics from both per-tensor and per-channel perspectives, and propose a Bimodal Integration strategy, which utilizes a mathematically equivalent sign operation to transform the bimodal distribution into a relatively easy-quantized normal distribution offline. Second, SAM encompasses diverse attention mechanisms (i.e., self-attention and two-way cross-attention), resulting in substantial variations in the post-Softmax distributions. Therefore, we introduce an Adaptive Granularity Quantization for Softmax through searching the optimal power-of-two base, which is hardware-friendly. Extensive experimental results across various vision tasks (instance segmentation, semantic segmentation and object detection), datasets and model variants show the superiority of PTQ4SAM. For example, when quantizing SAM-L to 6-bit, we achieve lossless accuracy for instance segmentation, about 0.5\% drop with theoretical 3.9$\times$ acceleration. The code is available at \url{https://github.com/chengtao-lv/PTQ4SAM}.

cs.CV↗

arXiv2024

Joint Chance-constrained Game for Coordinating Renewable Microgrids with Service Delivery Risk: A Bayesian Optimization Approach

Microgrids incorporate distributed energy resources (DERs) and flexible loads, which can provide energy and reserve services for the main grid. However, due to uncertain renewable generations such as solar power, microgrids might under-deliver reserve services and breach day-ahead contracts in real-time. If multiple microgrids breach their reserve contracts simultaneously, this could lead to a severe grid contingency. This paper designs a distributionally robust joint chance-constrained (DRJCC) game-theoretical framework considering uncertain real-time reserve provisions and the value of lost load (VoLL). Leveraging historical error samples, the reserve bidding strategy of each microgrid is formulated into a two-stage Wasserstein-metrics distribution robust optimization (DRO) model. A JCC is employed to regulate the under-delivered reserve capacity of all microgrids in a non-cooperative game. Considering the unknown correlation among players, a novel Bayesian optimization method approximates the optimal individual violation rates of microgrids and market equilibrium. The proposed game framework with the optimal rates is simulated with up to 14 players in a 30-bus network. Case studies are conducted using the California power market data. The proposed Bayesian method can effectively regulate the joint violation rate of the under-delivered reserve and secure the profit of microgrids in the reserve market.

math.OC↗