arXiv subjects

Zhican Wang

Publications and source records attributed to Zhican Wang.

7 recordsLinked to original sources

arXiv2026

Model Guides You How to Draw: Adaptive Visual Gating for Unified Multimodal Reasoning

Unified multimodal models (UMMs) with interleaved reasoning, which generate both textual and visual steps as part of intermediate reasoning traces, have demonstrated great potential for visual mathematical reasoning tasks. However, we identify a key insight in this paradigm: generating intermediate visual reasoning steps is not always beneficial and can even be harmful, as self-generated visual steps may introduce erroneous visual evidence that misleads subsequent reasoning. Moreover, frequently triggering visual steps during reasoning incurs substantial computational and memory overhead, degrading inference efficiency. To address these accuracy and efficiency challenges, we observe that the model's internal signals can indicate whether a visual step will benefit reasoning before the entire visual generation is completed. Specifically, this work identifies two internal signals: 1) Generation Intent, which reflects whether the model has a concrete textual plan for what to draw, and 2) Visual Fidelity, which measures whether the visual generation remains grounded in the original input image. Leveraging these internal signals, we propose AdaViG, a training-free adaptive visual gating method for unified multimodal reasoning. AdaViG dynamically evaluates each triggered visual step at an early visual generation stage and aborts it when both signals are weak, thereby preventing misleading visual evidence from entering the reasoning trace while avoiding unnecessary computation. Comprehensive experiments demonstrate that AdaViG improves accuracy by up to 5.7% while reducing visual generation FLOPs by 25.0%-91.0% and wall-clock latency by 15.4%-45.6%.

cs.CV↗

arXiv2026

Efficient 3D Gaussian Splatting with Axis-Shared Rasterization and Order-independent Transmittance

3D Gaussian Splatting (3DGS) has emerged as a powerful technique for novel view synthesis, combining high-quality reconstruction with efficient rendering. It has been widely adopted in domains such as AR/VR, robotics, and autonomous driving. However, achieving real-time performance on resource-constrained platforms remains challenging due to strict power and area budgets. Prior accelerators improve hardware performance but still overlook key inefficiencies, including insufficient rasterization efficiency, poor sorting scalability, and pipeline imbalance. This paper presents an architecture-algorithm co-design to address these challenges. First, we propose axis-shared rasterization, which precomputes and reuses common terms along the X- and Y-axes, reducing multiply-and-accumulate (MAC) operations by up to 38% while preserving high parallelism. Second, we develop a novel order-independent transmittance method that removes the need for explicit sorting by leveraging a lightweight multilayer perceptron (MLP) to directly approximate the transmittance of each Gaussian, enabling efficient alpha blending with negligible quality loss. Third, we design a unified reconfigurable PE array that supports both rasterization and MLP inference, sustaining high utilization without costly sorting hardware. Our experiments demonstrate that our design preserves rendering quality while achieving a 1.33 to 1.88x speedup over state-of-the-art 3DGS accelerators. Our code is open source at https://github.com/WangZhican/ISCA26_3DGS_Acc.

cs.GR↗

arXiv2026

AdaBlock-dLLM: Semantic-Aware Diffusion LLM Inference via Adaptive Block Size

Diffusion-based large language models (dLLMs) are gaining attention for their inherent capacity for parallel decoding, offering a compelling alternative to autoregressive LLMs. Among various decoding strategies, block-wise semi-autoregressive (semi-AR) approaches are widely adopted due to their support for KV caching and their favorable accuracy-speed trade-off. However, this paper identifies two fundamental limitations in the conventional semi-AR decoding approach that applies a fixed block size: i) late decoding overhead, where the unmasking of high-confidence tokens outside the current block is unnecessarily delayed, and ii) premature decoding error, where low-confidence tokens inside the current block are committed too early, leading to incorrect tokens. This paper presents the first systematic investigation challenging the fixed block size setting in semi-AR decoding. Through a statistical analysis of confidence dynamics during the denoising process, we identify a volatility band (VB) region during dLLM decoding, which encodes local semantic structure and can be used to guide adaptive block sizing. Leveraging these insights, we introduce AdaBlock-dLLM, a training-free, plug-and-play scheduler that adaptively aligns block boundaries with semantic steps by adjusting block size during runtime. Extensive experiments across diverse benchmarks show that AdaBlock-dLLM achieves up to 5.3% accuracy improvement under the same throughput budget. Beyond inference-time optimization, we hope our semantics-aware adaptive scheduling approach and confidence-based analysis will inspire future training strategies for dLLMs. Our code is available at https://github.com/lgxi24/AdaBlock-dLLM.

cs.LG↗

arXiv2025

SD-Acc: Accelerating Stable Diffusion through Phase-aware Sampling and Hardware Co-Optimizations

The emergence of diffusion models has significantly advanced generative AI, improving the quality, realism, and creativity of image and video generation. Among them, Stable Diffusion (StableDiff) stands out as a key model for text-to-image generation and a foundation for next-generation multi-modal algorithms. However, its high computational and memory demands hinder inference speed and energy efficiency. To address these challenges, we identify three core issues: (1) intensive and often redundant computations, (2) heterogeneous operations involving convolutions and attention mechanisms, and (3) diverse weight and activation sizes. We present SD-Acc, a novel algorithm and hardware co-optimization framework. At the algorithm level, we observe that high-level features in certain denoising phases show significant similarity, enabling approximate computation. Leveraging this, we propose an adaptive, phase-aware sampling strategy that reduces compute and memory loads. This framework automatically balances image quality and complexity based on the StableDiff model and user requirements. At the hardware level, we design an address-centric dataflow to efficiently handle heterogeneous operations within a simple systolic array. We address the bottleneck of nonlinear functions via a two-stage streaming architecture and a reconfigurable vector processing unit. Additionally, we implement adaptive dataflow optimizations by combining dynamic reuse and operator fusion tailored to StableDiff workloads, significantly reducing memory access. Across multiple StableDiff models, our method achieves up to a 3x reduction in computational demand without compromising image quality. Combined with our optimized hardware accelerator, SD-Acc delivers higher speed and energy efficiency than traditional CPU and GPU implementations.

cs.AR↗

arXiv2025

VEDA: Efficient LLM Generation Through Voting-based KV Cache Eviction and Dataflow-flexible Accelerator

Large Language Models (LLMs) excel in natural language processing tasks but pose significant computational and memory challenges for edge deployment due to their intensive resource demands. This work addresses the efficiency of LLM inference by algorithm-hardware-dataflow tri-optimizations. We propose a novel voting-based KV cache eviction algorithm, balancing hardware efficiency and algorithm accuracy by adaptively identifying unimportant kv vectors. From a dataflow perspective, we introduce a flexible-product dataflow and a runtime reconfigurable PE array for matrix-vector multiplication. The proposed approach effectively handles the diverse dimensional requirements and solves the challenges of incrementally varying sequence lengths. Additionally, an element-serial scheduling scheme is proposed for nonlinear operations, such as softmax and layer normalization (layernorm). Results demonstrate a substantial reduction in latency, accompanied by a significant decrease in hardware complexity, from O(N) to O(1). The proposed solution is realized in a custom-designed accelerator, VEDA, which outperforms existing hardware platforms. This research represents a significant advancement in LLM inference on resource-constrained edge devices, facilitating real-time processing, enhancing data privacy, and enabling model customization.

cs.AR↗

arXiv2025

Exploring Code Language Models for Automated HLS-based Hardware Generation: Benchmark, Infrastructure and Analysis

Recent advances in code generation have illuminated the potential of employing large language models (LLMs) for general-purpose programming languages such as Python and C++, opening new opportunities for automating software development and enhancing programmer productivity. The potential of LLMs in software programming has sparked significant interest in exploring automated hardware generation and automation. Although preliminary endeavors have been made to adopt LLMs in generating hardware description languages (HDLs), several challenges persist in this direction. First, the volume of available HDL training data is substantially smaller compared to that for software programming languages. Second, the pre-trained LLMs, mainly tailored for software code, tend to produce HDL designs that are more error-prone. Third, the generation of HDL requires a significantly higher number of tokens compared to software programming, leading to inefficiencies in cost and energy consumption. To tackle these challenges, this paper explores leveraging LLMs to generate High-Level Synthesis (HLS)-based hardware design. Although code generation for domain-specific programming languages is not new in the literature, we aim to provide experimental results, insights, benchmarks, and evaluation infrastructure to investigate the suitability of HLS over low-level HDLs for LLM-assisted hardware design generation. To achieve this, we first finetune pre-trained models for HLS-based hardware generation, using a collected dataset with text prompts and corresponding reference HLS designs. An LLM-assisted framework is then proposed to automate end-to-end hardware code generation, which also investigates the impact of chain-of-thought and feedback loops promoting techniques on HLS-design generation. Limited by the timeframe of this research, we plan to evaluate more advanced reasoning models in the future.

cs.LG↗

arXiv2024

DEFA: Efficient Deformable Attention Acceleration via Pruning-Assisted Grid-Sampling and Multi-Scale Parallel Processing

Multi-scale deformable attention (MSDeformAttn) has emerged as a key mechanism in various vision tasks, demonstrating explicit superiority attributed to multi-scale grid-sampling. However, this newly introduced operator incurs irregular data access and enormous memory requirement, leading to severe PE underutilization. Meanwhile, existing approaches for attention acceleration cannot be directly applied to MSDeformAttn due to lack of support for this distinct procedure. Therefore, we propose a dedicated algorithm-architecture co-design dubbed DEFA, the first-of-its-kind method for MSDeformAttn acceleration. At the algorithm level, DEFA adopts frequency-weighted pruning and probability-aware pruning for feature maps and sampling points respectively, alleviating the memory footprint by over 80%. At the architecture level, it explores the multi-scale parallelism to boost the throughput significantly and further reduces the memory access via fine-grained layer fusion and feature map reusing. Extensively evaluated on representative benchmarks, DEFA achieves 10.1-31.9x speedup and 20.3-37.7x energy efficiency boost compared to powerful GPUs. It also rivals the related accelerators by 2.2-3.7x energy efficiency improvement while providing pioneering support for MSDeformAttn.

cs.AR↗