arXiv subjects

Chiyue Wei

Publications and source records attributed to Chiyue Wei.

16 recordsLinked to original sources

arXiv2026

Optimus: Elastic Decoding for Efficient Diffusion LLM Serving

Large language model (LLM) serving is fundamentally limited by inefficient hardware utilization. Autoregressive (AR) decoding underutilizes GPUs due to its strictly sequential execution, while diffusion LLMs (DLLMs) improve throughput by decoding multiple tokens per iteration. However, fixed block-size diffusion decoding exhibits strong load sensitivity: large blocks exploit idle GPU resources under low load, but saturate early and incur substantial redundant computation under high load. As a result, throughput gains vanish beyond saturation, and no single decoding granularity performs well across dynamic serving workloads. We present Optimus, a serving system that enables elastic decoding for diffusion LLMs by dynamically adapting decoding granularity to runtime load. The key idea is to treat decoding granularity as a runtime control variable, balancing GPU utilization and token efficiency. Optimus combines chunked decoding, which enables fine-grained execution without retraining, with saturation-aware scheduling, a closed-loop mechanism that selects chunk sizes based on runtime conditions. Together with system-level optimizations and customized attention kernels, Optimus achieves significant performance improvements while preserving model accuracy. Experiments show that Optimus delivers up to 6.1x throughput improvement over AR decoding and 4.3x improvement over fixed-block diffusion LLM, while maintaining stable performance across diverse load regimes and improving end-to-end serving capacity under latency constraints. The source code is available at https://github.com/dubcyfor3/Optimus.

cs.DC↗

arXiv2026

EVA: Accelerating LLM Decoding via an Efficient Vector Quantization Architecture

Large Language Models (LLMs) have achieved impressive performance across diverse domains but remain inefficient during the autoregressive decoding phase. Unlike the prefill stage, which employs compute-bound GEMM operations, decoding executes a sequence of small GEMV-like computations that are memory-bound and underutilize modern accelerators. Weight-only vector quantization (VQ) has emerged as an effective compression technique that clusters model weights into a shared codebook and replaces the original weight matrix with low-precision indices, enabling 2-bit-level weight compression. While this approach substantially reduces model size and memory bandwidth, it still suffers from two critical inefficiencies: the low utilization of GEMV computation and frequent memory conflicts during codebook lookups. This paper presents EVA, an efficient vector-quantization-based architecture that addresses both computational and memory bottlenecks in LLM decoding. EVA builds on a simple yet effective insight that combines input-codebook computation with conflict-free memory access. Instead of reconstructing quantized weights from indices, EVA directly performs dot products between input vectors and the weight codebook, transforming LLM decoding from GEMV to GEMM computation. It then performs structured lookups from an intermediate output buffer, eliminating memory bank conflicts. We further design a hardware-software co-optimized architecture specialized for LLM decoding while remaining compatible with conventional prefill execution. Evaluations show that EVA achieves up to 11.17$\times$ speedup and 7.17$\times$ higher energy efficiency compared with the SOTA lookup-based architecture, while preserving arithmetic precision after vector quantization. Our code is available at https://github.com/dbw6/Eva.git.

cs.AR↗

arXiv2026

MARS: Efficient, Adaptive Co-Scheduling for Heterogeneous Agentic Systems

Large language models (LLMs) are increasingly deployed as the execution core of autonomous agents rather than as standalone text generators. Agentic workloads induce a temporal shift from single-turn inference to multi-turn LLM-tool loops, and a spatial shift from chat-scale, GPU-only execution to repository-scale, GPU-CPU co-located execution. Consequently, coordinating heterogeneous resource demands of agentic execution has emerged as a critical system challenge. We design and implement MARS, an efficient and adaptive co-scheduling system that globally coordinates heterogeneous agentic workloads under coupled GPU-CPU resource pressure. By establishing holistic visibility across GPU inference and CPU tool execution via a unified information stream, an external control plane in MARS decouples admission from execution to prevent heterogeneous resource oversubscription. An internal agent-centric scheduler further minimizes the end-to-end critical path by prioritizing latency-sensitive continuations and adaptively retaining KV cache state only when warm resumption yields a latency benefit. Our evaluations show that MARS reduces end-to-end latency by up to 5.94x while maintaining nearly maximal system throughput. We further integrate MARS as the serving backend for the OpenHands coding agent framework, demonstrating its real-world effectiveness by accelerating end-to-end task completion time by up to 1.87x. Our source code is publicly available at https://github.com/Afterglow231/MARS_preview .

cs.OS↗

arXiv2026

T2S-Bench & Structure-of-Thought: Benchmarking and Prompting Comprehensive Text-to-Structure Reasoning

Think about how human handles complex reading tasks: marking key points, inferring their relationships, and structuring information to guide understanding and responses. Likewise, can a large language model benefit from text structure to enhance text-processing performance? To explore it, in this work, we first introduce Structure of Thought (SoT), a prompting technique that explicitly guides models to construct intermediate text structures, consistently boosting performance across eight tasks and three model families. Building upon this insight, we present T2S-Bench, the first benchmark designed to evaluate and improve text-to-structure capabilities of models. T2S-Bench includes 1.8K samples across 6 scientific domains and 32 structural types, rigorously constructed to ensure accuracy, fairness, and quality. Evaluation on 45 mainstream models reveals substantial improvement potential: the average accuracy on the multi-hop reasoning task is only 52.1%, and even the most advanced model achieves 58.1% node accuracy in end-to-end extraction. Furthermore, on Qwen2.5-7B-Instruct, SoT alone yields an average +5.7% improvement across eight diverse text-processing tasks, and fine-tuning on T2S-Bench further increases this gain to +8.6%. These results highlight the value of explicit text structuring and the complementary contributions of SoT and T2S-Bench. Dataset and eval code have been released at https://t2s-bench.github.io/T2S-Bench-Page/.

cs.CL↗

arXiv2025

Focus: A Streaming Concentration Architecture for Efficient Vision-Language Models

Vision-Language Models (VLMs) have demonstrated strong performance on tasks such as video captioning and visual question answering. However, their growing scale and video-level inputs lead to significant computational and memory overhead, posing challenges for real-time deployment on hardware accelerators. While prior work attempts to reduce redundancy via token pruning or merging, these methods typically operate at coarse granularity and incur high runtime overhead due to global token-level operations. In this study, we propose Focus, a Streaming Concentration Architecture that efficiently accelerates VLM inference through progressive, fine-grained redundancy elimination. Focus introduces a multilevel concentration paradigm that hierarchically compresses vision-language inputs at three levels: (1) semantic-guided token pruning based on textual prompts, (2) spatial-temporal block-level concentration using localized comparisons, and (3) vector-level redundancy removal via motion-aware matching. All concentration steps are tightly co-designed with the architecture to support streaming-friendly, on-chip execution. Focus leverages GEMM tiling, convolution-style layout, and cross-modal attention to minimize off-chip access while enabling high throughput. Implemented as a modular unit within a systolic-array accelerator, Focus achieves a 2.4x speedup and 3.3x reduction in energy, significantly outperforming state-of-the-art accelerators in both performance and energy efficiency. Full-stack implementation of Focus is open-sourced at https://github.com/dubcyfor3/Focus.

cs.AR↗

arXiv2025

FractalCloud: A Fractal-Inspired Architecture for Efficient Large-Scale Point Cloud Processing

Three-dimensional (3D) point clouds are increasingly used in applications such as autonomous driving, robotics, and virtual reality (VR). Point-based neural networks (PNNs) have demonstrated strong performance in point cloud analysis, originally targeting small-scale inputs. However, as PNNs evolve to process large-scale point clouds with hundreds of thousands of points, all-to-all computation and global memory access in point cloud processing introduce substantial overhead, causing $O(n^2)$ computational complexity and memory traffic where n is the number of points}. Existing accelerators, primarily optimized for small-scale workloads, overlook this challenge and scale poorly due to inefficient partitioning and non-parallel architectures. To address these issues, we propose FractalCloud, a fractal-inspired hardware architecture for efficient large-scale 3D point cloud processing. FractalCloud introduces two key optimizations: (1) a co-designed Fractal method for shape-aware and hardware-friendly partitioning, and (2) block-parallel point operations that decompose and parallelize all point operations. A dedicated hardware design with on-chip fractal and flexible parallelism further enables fully parallel processing within limited memory resources. Implemented in 28 nm technology as a chip layout with a core area of 1.5 $mm^2$, FractalCloud achieves 21.7x speedup and 27x energy reduction over state-of-the-art accelerators while maintaining network accuracy, demonstrating its scalability and efficiency for PNN inference.

cs.AR↗

arXiv2025

Platinum: Path-Adaptable LUT-Based Accelerator Tailored for Low-Bit Weight Matrix Multiplication

The rapid scaling of large language models demands more efficient hardware. Quantization offers a promising trade-off between efficiency and performance. With ultra-low-bit quantization, there are abundant opportunities for results reuse, and thus it can be boosted with lookup tables (LUTs) based acceleration. However, existing LUT-based methods suffer from computation and hardware overheads for LUT construction, and rely solely on bit-serial computation, which is suboptimal for ternary-weight networks. We propose Platinum, a lightweight ASIC accelerator for integer weight mixed-precision matrix multiplication (mpGEMM) using LUTs. Platinum reduces LUT construction overhead via offline-generated construction paths and supports both general bit-serial and optimized ternary-weight execution through adaptive path switching. On BitNet b1.58-3B, Platinum achieves up to 73.6x, 4.09x, and 2.15x speedups over SpikingEyeriss, Prosperity, and 16-thread T-MAC (CPU), respectively, along with energy reductions of 32.4x, 3.23x, and 20.9x, all within a 0.96mm2 chip area. This demonstrates the potential of LUT-based ASICs as efficient, scalable solutions for ultra-low-bit neural networks on edge platforms.

cs.AR↗

arXiv2025

CAMformer: Associative Memory is All You Need

Transformers face scalability challenges due to the quadratic cost of attention, which involves dense similarity computations between queries and keys. We propose CAMformer, a novel accelerator that reinterprets attention as an associative memory operation and computes attention scores using a voltage-domain Binary Attention Content Addressable Memory (BA-CAM). This enables constant-time similarity search through analog charge sharing, replacing digital arithmetic with physical similarity sensing. CAMformer integrates hierarchical two-stage top-k filtering, pipelined execution, and high-precision contextualization to achieve both algorithmic accuracy and architectural efficiency. Evaluated on BERT and Vision Transformer workloads, CAMformer achieves over 10x energy efficiency, up to 4x higher throughput, and 6-8x lower area compared to state-of-the-art accelerators--while maintaining near-lossless accuracy.

cs.AR↗

arXiv2025

DPad: Efficient Diffusion Language Models with Suffix Dropout

Diffusion-based Large Language Models (dLLMs) parallelize text generation by framing decoding as a denoising process, but suffer from high computational overhead since they predict all future suffix tokens at each step while retaining only a small fraction. We propose Diffusion Scratchpad (DPad), a training-free method that restricts attention to a small set of nearby suffix tokens, preserving fidelity while eliminating redundancy. DPad integrates two strategies: (i) a sliding window, which maintains a fixed-length suffix window, and (ii) distance-decay dropout, which deterministically removes distant suffix tokens before attention computation. This simple design is compatible with existing optimizations such as prefix caching and can be implemented with only a few lines of code. Comprehensive evaluations across multiple benchmarks on LLaDA-1.5 and Dream models demonstrate that DPad delivers up to $\mathbf{61.4\times}$ speedup over vanilla dLLMs while maintaining comparable accuracy, highlighting its potential for efficient and scalable long-sequence inference. Our code is available at https://github.com/Crys-Chen/DPad.

cs.CL↗

arXiv2025

Phi: Leveraging Pattern-based Hierarchical Sparsity for High-Efficiency Spiking Neural Networks

Spiking Neural Networks (SNNs) are gaining attention for their energy efficiency and biological plausibility, utilizing 0-1 activation sparsity through spike-driven computation. While existing SNN accelerators exploit this sparsity to skip zero computations, they often overlook the unique distribution patterns inherent in binary activations. In this work, we observe that particular patterns exist in spike activations, which we can utilize to reduce the substantial computation of SNN models. Based on these findings, we propose a novel \textbf{pattern-based hierarchical sparsity} framework, termed \textbf{\textit{Phi}}, to optimize computation. \textit{Phi} introduces a two-level sparsity hierarchy: Level 1 exhibits vector-wise sparsity by representing activations with pre-defined patterns, allowing for offline pre-computation with weights and significantly reducing most runtime computation. Level 2 features element-wise sparsity by complementing the Level 1 matrix, using a highly sparse matrix to further reduce computation while maintaining accuracy. We present an algorithm-hardware co-design approach. Algorithmically, we employ a k-means-based pattern selection method to identify representative patterns and introduce a pattern-aware fine-tuning technique to enhance Level 2 sparsity. Architecturally, we design \textbf{\textit{Phi}}, a dedicated hardware architecture that efficiently processes the two levels of \textit{Phi} sparsity on the fly. Extensive experiments demonstrate that \textit{Phi} achieves a $3.45\times$ speedup and a $4.93\times$ improvement in energy efficiency compared to state-of-the-art SNN accelerators, showcasing the effectiveness of our framework in optimizing SNN computation.

cs.AR↗

arXiv2025

Ecco: Improving Memory Bandwidth and Capacity for LLMs via Entropy-aware Cache Compression

Large language models (LLMs) have demonstrated transformative capabilities across diverse artificial intelligence applications, yet their deployment is hindered by substantial memory and computational demands, especially in resource-constrained environments. Quantization techniques have emerged as a critical solution, reducing data precision to enhance memory and computational efficiency. However, existing methods often suffer from high runtime overheads and potential accuracy degradation. To address these challenges, we propose Ecco, an entropy-based cache compression technique tailored for LLMs. Ecco combines group-wise and non-uniform quantization with pre-defined shared k-means patterns and Huffman coding to exploit the inherent entropy characteristics of LLM cache data. Recognizing the inefficiencies of traditional Huffman coding in terms of parallelism and latency, we introduce a novel parallel Huffman-based decoding process with a multi-stage pipeline design, reducing latency by two orders of magnitude and achieving throughput comparable to GPU L2 caches. Comprehensive evaluations demonstrate that Ecco achieves an up to 2.9$\times$ and 1.9$\times$ speedup over the state-of-the-art AWQ and SmoothQuant framework, 2.4$\times$ over the Olive accelerator, all while increasing memory capacity by nearly 4$\times$ and maintaining state-of-the-art LLM accuracy. These results underscore the effectiveness of our entropy-based cache compression in enhancing LLM performance and efficiency, paving the way for more deployable large-scale AI models.

cs.AR↗

arXiv2025

Transitive Array: An Efficient GEMM Accelerator with Result Reuse

Deep Neural Networks (DNNs) and Large Language Models (LLMs) have revolutionized artificial intelligence, yet their deployment faces significant memory and computational challenges, especially in resource-constrained environments. Quantization techniques have mitigated some of these issues by reducing data precision, primarily focusing on General Matrix Multiplication (GEMM). This study introduces a novel sparsity paradigm, transitive sparsity, which leverages the reuse of previously computed results to substantially minimize computational overhead in GEMM operations. By representing transitive relations using a directed acyclic graph, we develop an efficient strategy for determining optimal execution orders, thereby overcoming inherent challenges related to execution dependencies and parallelism. Building on this foundation, we present the Transitive Array, a multiplication-free accelerator designed to exploit transitive sparsity in GEMM. Our architecture effectively balances computational workloads across multiple parallel lanes, ensuring high efficiency and optimal resource utilization. Comprehensive evaluations demonstrate that the Transitive Array achieves approximately 7.46$\times$ and 3.97$\times$ speedup and 2.31$\times$ and 1.65$\times$ energy reduction compared to state-of-the-art accelerators such as Olive and BitVert while maintaining comparable model accuracy on LLaMA models.

cs.AR↗

arXiv2025

Prosperity: Accelerating Spiking Neural Networks via Product Sparsity

Spiking Neural Networks (SNNs) are highly efficient due to their spike-based activation, which inherently produces bit-sparse computation patterns. Existing hardware implementations of SNNs leverage this sparsity pattern to avoid wasteful zero-value computations, yet this approach fails to fully capitalize on the potential efficiency of SNNs. This study introduces a novel sparsity paradigm called Product Sparsity, which leverages combinatorial similarities within matrix multiplication operations to reuse the inner product result and reduce redundant computations. Product Sparsity significantly enhances sparsity in SNNs without compromising the original computation results compared to traditional bit sparsity methods. For instance, in the SpikeBERT SNN model, Product Sparsity achieves a density of only $1.23\%$ and reduces computation by $11\times$, compared to bit sparsity, which has a density of $13.19\%$. To efficiently implement Product Sparsity, we propose Prosperity, an architecture that addresses the challenges of identifying and eliminating redundant computations in real-time. Compared to prior SNN accelerator PTB and the A100 GPU, Prosperity achieves an average speedup of $7.4\times$ and $1.8\times$, respectively, along with energy efficiency improvements of $8.0\times$ and $193\times$, respectively. The code for Prosperity is available at https://github.com/dubcyfor3/Prosperity.

cs.AR↗

arXiv2025

Hamming Attention Distillation: Binarizing Keys and Queries for Efficient Long-Context Transformers

Pre-trained transformer models with extended context windows are notoriously expensive to run at scale, often limiting real-world deployment due to their high computational and memory requirements. In this paper, we introduce Hamming Attention Distillation (HAD), a novel framework that binarizes keys and queries in the attention mechanism to achieve significant efficiency gains. By converting keys and queries into {-1, +1} vectors and replacing dot-product operations with efficient Hamming distance computations, our method drastically reduces computational overhead. Additionally, we incorporate attention matrix sparsification to prune low-impact activations, which further reduces the cost of processing long-context sequences. \par Despite these aggressive compression strategies, our distilled approach preserves a high degree of representational power, leading to substantially improved accuracy compared to prior transformer binarization methods. We evaluate HAD on a range of tasks and models, including the GLUE benchmark, ImageNet, and QuALITY, demonstrating state-of-the-art performance among binarized Transformers while drastically reducing the computational costs of long-context inference. \par We implement HAD in custom hardware simulations, demonstrating superior performance characteristics compared to a custom hardware implementation of standard attention. HAD achieves just $\mathbf{1.78}\%$ performance losses on GLUE compared to $9.08\%$ in state-of-the-art binarization work, and $\mathbf{2.5}\%$ performance losses on ImageNet compared to $12.14\%$, all while targeting custom hardware with a $\mathbf{79}\%$ area reduction and $\mathbf{87}\%$ power reduction compared to its standard attention counterpart.

cs.LG↗

arXiv2024

A Survey: Collaborative Hardware and Software Design in the Era of Large Language Models

The rapid development of large language models (LLMs) has significantly transformed the field of artificial intelligence, demonstrating remarkable capabilities in natural language processing and moving towards multi-modal functionality. These models are increasingly integrated into diverse applications, impacting both research and industry. However, their development and deployment present substantial challenges, including the need for extensive computational resources, high energy consumption, and complex software optimizations. Unlike traditional deep learning systems, LLMs require unique optimization strategies for training and inference, focusing on system-level efficiency. This paper surveys hardware and software co-design approaches specifically tailored to address the unique characteristics and constraints of large language models. This survey analyzes the challenges and impacts of LLMs on hardware and algorithm research, exploring algorithm optimization, hardware design, and system-level innovations. It aims to provide a comprehensive understanding of the trade-offs and considerations in LLM-centric computing systems, guiding future advancements in AI. Finally, we summarize the existing efforts in this space and outline future directions toward realizing production-grade co-design methodologies for the next generation of large language models and AI systems.

cs.AR↗

arXiv2023

DeSCo: Towards Generalizable and Scalable Deep Subgraph Counting

We introduce DeSCo, a scalable neural deep subgraph counting pipeline, designed to accurately predict both the count and occurrence position of queries on target graphs post single training. Firstly, DeSCo uses a novel canonical partition and divides the large target graph into small neighborhood graphs, greatly reducing the count variation while guaranteeing no missing or double-counting. Secondly, neighborhood counting uses an expressive subgraph-based heterogeneous graph neural network to accurately count in each neighborhood. Finally, gossip propagation propagates neighborhood counts with learnable gates to harness the inductive biases of motif counts. DeSCo is evaluated on eight real-world datasets from various domains. It outperforms state-of-the-art neural methods with 137x improvement in the mean squared error of count prediction, while maintaining the polynomial runtime complexity. Our open source project is at https://github.com/fuvty/DeSCo.

cs.LG↗