论文检索

Yang Yu, Peiyu Zang, Chi Hsu Tsai, Haiming Wu, Yixin Shen, Jialing Zhang, Haoyu Wang, Zhiyou Xiao, Jingze Shi, Yuyu Luo 等

The performance of modern AI systems is fundamentally constrained by the quality of their underlying GPU kernels, which translate high-level algorithmic semantics into low-level hardware operations. Achieving near-optimal kernels requires expert-level understanding of hardware architectures and programming models, making kernel engineering a critical but notoriously time-consuming and non-scalable process. Recent advances in large language models and LLM-based agents have opened new possibilities for automating kernel generation and optimization. LLMs are well-suited to compress expert-level kernel knowledge that is difficult to formalize, while agentic systems further enable scalable optimization by casting kernel development as an iterative, feedback-driven loop. Rapid progress has been made in this area. However, the field remains fragmented and lacks a systematic perspective for LLM-driven kernel generation. This survey addresses this gap by providing a structured overview of existing approaches, spanning LLM-based approaches and agentic optimization workflows, and systematically organizing the datasets and benchmarks that underpin learning and evaluation in this domain. Moreover, key open challenges and future research directions are further outlined, aiming to establish a comprehensive reference for the next generation of automated kernel optimization. To keep track of this field, we maintain an open-source GitHub repository at https://github.com/flagos-ai/awesome-LLM-driven-kernel-generation.

Yubang Wang, Chenxi Zhang, Bowen Chen, Zezheng Huai, Zihao Dai, Xinchi Chen, Yuxin Wang, Yining Zheng, Jingjing Gong, Xipeng Qiu

Autonomous agents are increasingly expected to support scientific research, and recent benchmarks report progress in code repair and autonomous experimentation. However, these evaluations typically assume a pre-configured execution environment, which requires resolving complex software dependencies, aligning hardware and framework versions, and configuring distributed execution, yet this capability remains largely unbenchmarked. We introduce ResearchEnvBench, a benchmark for environment synthesis in research code execution. Given a research repository, documentation, and a target execution setting, agents must construct an environment that successfully executes at runtime. Evaluations on diverse research repositories reveal a substantial gap in current SOTA agents, with failures dominated by incomplete dependency resolution and brittle version coupling. ResearchEnvBench provides a realistic testbed for advancing autonomous agents toward reproducible scientific research.

Yuze Sun, Wentao Luo, Yanfei Xiang, Jiancheng Pan, Jiahao Li, Quan Zhang, Xiaomeng Huang

With the growing role of artificial intelligence in climate and weather research, efficient model training and inference are in high demand. Current models like FourCastNet and AI-GOMS depend heavily on GPUs, limiting hardware independence, especially for Chinese domestic hardware and frameworks. To address this issue, we present a framework for migrating large-scale atmospheric and oceanic models from PyTorch to MindSpore and optimizing for Chinese chips, and evaluating their performance against GPUs. The framework focuses on software-hardware adaptation, memory optimization, and parallelism. Furthermore, the model's performance is evaluated across multiple metrics, including training speed, inference speed, model accuracy, and energy efficiency, with comparisons against GPU-based implementations. Experimental results demonstrate that the migration and optimization process preserves the models' original accuracy while significantly reducing system dependencies and improving operational efficiency by leveraging Chinese chips as a viable alternative for scientific computing. This work provides valuable insights and practical guidance for leveraging Chinese domestic chips and frameworks in atmospheric and oceanic AI model development, offering a pathway toward greater technological independence.

Dejan Milojevic, Gioele Zardini, Miriam Elser, Andrea Censi, Emilio Frazzoli

This article discusses the integration challenges and strategies for designing mobile robots, by focusing on the task-driven, optimal selection of hardware and software to balance safety, efficiency, and minimal usage of resources such as costs, energy, computational requirements, and weight. We emphasize the interplay between perception and motion planning in decision-making by introducing the concept of occupancy queries to quantify the perception requirements for sampling-based motion planners. Sensor and algorithm performance are evaluated using false negative rate and false positive rate across various factors such as geometric relationships, object properties, sensor resolution, and environmental conditions. By integrating perception requirements with perception performance, an integer linear programming approach is proposed for efficient sensor and algorithm selection and placement. This forms the basis for a co-design optimization that includes the robot body, motion planner, perception pipeline, and computing unit. We refer to this framework for solving the co-design problem of mobile robots as CODEI, short for co-design of embodied intelligence. A case study on developing an autonomous vehicle for urban scenarios provides actionable information for designers, and shows that complex tasks escalate resource demands, with task performance affecting choices of the autonomy stack. The study demonstrates that resource prioritization influences sensor choice: cameras are preferred for cost-effective and lightweight designs, while lidar sensors are chosen for better energy and computational efficiency.

You Li, Guannan Zhao, Yuhao Ju, Yunqi He, Jie Gu, Hai Zhou

We introduce LLA, an effective intellectual property (IP) protection scheme for generative AI models. LLA leverages the synergy between hardware and software to defend against various supply chain threats, including model theft, model corruption, and information leakage. On the software side, it embeds key bits into neurons that can trigger outliers to degrade performance and applies invariance transformations to obscure the key values. On the hardware side, it integrates a lightweight locking module into the AI accelerator while maintaining compatibility with various dataflow patterns and toolchains. An accelerator with a pre-stored secret key acts as a license to access the model services provided by the IP owner. The evaluation results show that LLA can withstand a broad range of oracle-guided key optimization attacks, while incurring a minimal computational overhead of less than 0.1% for 7,168 key bits.

Ge Wang, Xing Liu, Xin Yuan

Video snapshot compressive imaging (SCI) captures dynamic scene sequences through a two-dimensional (2D) snapshot, fundamentally relying on optical modulation for hardware compression and the corresponding software reconstruction. While mainstream video SCI using random binary modulation has demonstrated success, it inevitably results in temporal aliasing during compression. One-hot modulation, activating only one sub-frame per pixel, provides a promising solution for achieving perfect temporal decoupling, thereby alleviating issues associated with aliasing. However, no algorithms currently exist to fully exploit this potential. To bridge this gap, we propose an algorithm specifically designed for one-hot masks. First, leveraging the decoupling properties of one-hot modulation, we transform the reconstruction task into a generative video inpainting problem and introduce a stochastic differential equation (SDE) of the forward process that aligns with the hardware compression process. Next, we identify limitations of the pure diffusion method for video SCI and propose a novel framework that combines one-step regression initialization with one-step diffusion refinement. Furthermore, to mitigate the spatial degradation caused by one-hot modulation, we implement a dual optical path at the hardware level, utilizing complementary information from another path to enhance the inpainted video. To our knowledge, this is the first work integrating diffusion into video SCI reconstruction. Experiments conducted on synthetic datasets and real scenes demonstrate the effectiveness of our method.

Tomoya Tsuchida, Keita Takahashi, Chihiro Tsutake, Toshiaki Fujii, Hajime Nagahara

We propose Coded-E2LF (coded event to light field), a computational imaging method for acquiring a 4-D light field using a coded aperture and a stationary event-only camera. In a previous work, an imaging system similar to ours was adopted, but both events and intensity images were captured and used for light field reconstruction. In contrast, our method is purely event-based, which relaxes restrictions for hardware implementation. We also introduce several advancements from the previous work that enable us to theoretically support and practically improve light field reconstruction from events alone. In particular, we clarify the key role of a black pattern in aperture coding patterns. We finally implemented our method on real imaging hardware to demonstrate its effectiveness in capturing real 3-D scenes. To the best of our knowledge, we are the first to demonstrate that a 4-D light field with pixel-level accuracy can be reconstructed from events alone. Our software and supplementary video are available from our project website: https://www.fujii.nuee.nagoya-u.ac.jp/Research/EventLF/.

Zain Shabeeb, Daniel Saeedi, Darin Tsui, Vida Jamali, Amirali Aghazadeh

Cryo-electron microscopy (cryo-EM) enables the atomic-resolution visualization of biomolecules; however, modern direct detectors generate data volumes that far exceed the available storage and transfer bandwidth, thereby constraining practical throughput. We introduce cryoSENSE, the computational realization of a hardware-software co-designed framework for compressive cryo-EM sensing and acquisition. We show that cryo-EM images of proteins lie on low-dimensional manifolds that can be independently represented using sparse priors in predefined bases and generative priors captured by a denoising diffusion model. cryoSENSE leverages these low-dimensional manifolds to enable faithful image reconstruction from spatial and Fourier-domain undersampled measurements while preserving downstream structural resolution. In experiments, cryoSENSE increases acquisition throughput by up to 2.5xwhile retaining the original 3D resolution, offering controllable trade-offs between the number of masked measurements and the level of downsampling. Sparse priors favor faithful reconstruction from Fourier-domain measurements and moderate compression, whereas generative diffusion priors achieve accurate recovery from pixel-domain measurements and more severe undersampling.

Jun Zhang, Yicheng Ji, Feiyang Ren, Yihang Li, Bowen Zeng, Zonghao Chen, Ke Chen, Lidan Shou, Gang Chen, Huan Li

Large Vision-Language Models (LVLMs) enable sophisticated reasoning over images and videos, yet their inference is hindered by a systemic efficiency barrier known as visual token dominance. This overhead is driven by a multi-regime interplay between high-resolution feature extraction, quadratic attention scaling, and memory bandwidth constraints. We present a systematic taxonomy of efficiency techniques structured around the inference lifecycle, consisting of encoding, prefilling, and decoding. Unlike prior reviews focused on isolated optimizations, we analyze the end-to-end pipeline to reveal how upstream decisions dictate downstream bottlenecks, covering compute-bound visual encoding, the intensive prefilling of massive contexts, and the ”visual memory wall” in bandwidth-bound decoding. By decoupling the efficiency landscape into the axes of shaping information density, managing long-context attention, and overcoming memory limits, this work provides a structured analysis of how isolated optimizations compose to navigate the trade-off between visual fidelity and system efficiency. The survey concludes by outlining four future frontiers supported by pilot empirical insights, including hybrid compression based on functional unit sensitivity, modality-aware decoding with relaxed verification, progressive state management for streaming continuity, and stage-disaggregated serving through hardware-algorithm co-design. The submitted software contains a snapshot of our literature repository, which is designed to be maintained as a living resource for the community.

General Machine Learning · Hardware and Software

Rajat Vadiraj Dwaraknath, Sungyoon Kim, Mert Pilanci

Sparse sketches such as the sparse Johnson–Lindenstrauss transform are a core primitive in randomized numerical linear algebra because they leverage random sparsity to reduce the arithmetic cost of sketching, while still offering strong approximation guarantees. Their random sparsity, however, is at odds with efficient implementations on modern GPUs, since it leads to irregular memory access patterns that degrade memory bandwidth utilization. Motivated by this tension, we pursue a sketch–kernel co-design approach: we design a new family of sparse sketches, BlockPerm-SJLT, whose sparsity structure is chosen to enable FlashSketch, a corresponding optimized CUDA kernel that implements these sketches efficiently. The design of BlockPerm-SJLT introduces a tunable parameter that explicitly trades off the tension between GPU-efficiency and sketching robustness. We provide theoretical guarantees for BlockPerm-SJLT under the oblivious subspace embedding (OSE) framework, and also analyze the effect of the tunable parameter on sketching quality. We empirically evaluate FlashSketch on standard RandNLA benchmarks, as well as an end-to-end ML data attribution pipeline called GraSS. FlashSketch pushes the Pareto frontier of sketching quality versus speed, across a range of regimes and tasks, and achieves a global geomean speedup of roughly $1.7 \times$ over the prior state-of-the-art GPU sketches.

Deep Learning · Sequential Models, Time series

Julien Brandoit, Arthur Fyon, Damien Ernst, Guillaume Drion

Sequence learning is dominated by Transformers and parallelizable recurrent neural networks such as state-space models, yet learning long-term dependencies remains challenging, and state-of-the-art designs trade power consumption for performance. The Bistable Memory Recurrent Unit (BMRU) was introduced to enable hardware–software co-design of ultra-low power RNNs: quantized states with hysteresis provide persistent memory while mapping directly to analog primitives. However, BMRU performance lags behind parallelizable RNNs on complex sequential tasks. In this paper, we identify gradient blocking during state updates as a key limitation and propose a cumulative update formulation that restores gradient flow while preserving persistent memory, creating skip-connections through time. This leads to the Cumulative Memory Recurrent Unit (CMRU) and its relaxed variant, the $\alpha$CMRU. Experiments show that the cumulative formulation dramatically improves convergence stability and reduces initialization sensitivity. The CMRU and $\alpha$CMRU match the performance of Linear Recurrent Units (LRUs) and minimal Gated Recurrent Units (minGRUs) on standard benchmarks at small model sizes, while the CMRU retains quantized states, persistent memory, and noise-resilient dynamics essential for analog implementation.

General Machine Learning · Hardware and Software

Seungheon Lee, Jeongmin Sun, Jaeyong Chung

Differentiable logic gate networks map learned models directly to gate-level circuits, enabling ultra-low-latency inference, yet their logic footprint often exceeds FPGA capacity budgets. Tightly fitting a trained model to a target FPGA requires a post-training mechanism to trade off network complexity and accuracy—analogous to pruning in standard neural networks. To this end, we introduce *unit tying*: a simplification that forces selected gates to constants (0 or 1), enabling constant propagation and downstream logic elimination. However, we observe that naively extending pruning criteria to logic networks is unreliable under such near-discrete modifications. We therefore propose a two-stage algorithm for unit tying: (i) a fast Gauss–Newton screening step under a teacher-referenced logit-distortion objective that constructs a high-recall overshoot set and (ii) a refinement step that corrects approximation and interaction-driven errors using a small number of finite-difference evaluations. On CIFAR-10 and MNIST, our method consistently improves the accuracy–area trade-off over common saliency baselines, yielding substantial post-synthesis LUT reductions of up to 48% on CIFAR-10 and 43% on MNIST, with modest accuracy degradation.

General Machine Learning · Hardware and Software

Haoran Zhang, Zhiyu Liang, ZUO Decheng, Hongzhi Wang

Multi-tenant multi-LoRA serving multiplexes many LoRA adapters on a single GPU under high utilization, where most device memory is reserved for the KV cache, leaving only a small residency budget $K$ for adapters. In this regime, adapter readiness is atomic: if an adapter is not device-resident, the engine must perform a mandatory fetch, stalling shared execution and amplifying tail latency system-wide. With only $K$ residency slots, we identify a LoRA residency cliff: once the active adapter working set exceeds $K$, time-to-first-token (TTFT) tail latency can exhibit a congestion collapse rather than smooth degradation. To tame this cliff, we propose CLIMB, a minimal ingress controller that enforces feasibility-first admission by queueing non-resident adapters outside the engine, prioritizing critical (VIP) traffic, and rotating background adapters via round-robin. On a cliff-inducing workload, CLIMB averts collapse, reducing VIP TTFT p99 from 38.7 s to 13.1 s at matched throughput (10.66 rps) by keeping VIP engine latency near 0.13 s and shifting the residual tail into explicit ingress queueing. Overall, CLIMB shifts fetch-induced stalls from inside the engine to managed ingress queues, mitigating tail amplification without throughput loss in the evaluated settings.

General Machine Learning · Hardware and Software

Qinhong Ma, Yulin Chen, Zhiwei Fan, Suzhen Wu, Bo Mao

Weightless Neural Networks (WNNs) based on interconnected Lookup Tables (LUTs) have attracted attention for inference in extremely compact models, but achieving competitive accuracy under such tight resource budgets remains challenging. To address these issues, we introduce the Convolutional Learnable-Group Weightless Neural Network (CLGN). CLGN constructs convolutional layers using LUTs and incorporates a learnable GroupSum connection, thereby enhancing the accuracy of WNNs while maintaining low implementation resource consumption. Moreover, we propose a hierarchical training strategy to improve the training efficiency. We evaluate CLGN in two edge computing scenarios: (1) FPGA, where we evaluate accuracy, latency, throughput, power consumption, LUTs usage, and parameter size; and (2) Microprocessor, where we evaluate latency and memory usage. Compared with the state-of-the-art solutions, the proposed CLGN achieves superior accuracy while maintaining lower implementation resource consumption.

General Machine Learning · Hardware and Software

Anselm Paulus, Andreas René Geist, Vit Musil, Sebastian Hoffmann, Georg Martius

Automatic differentiation (AD) frameworks such as JAX and PyTorch have enabled gradient-based optimization for a wide range of scientific fields. Yet, many ''hard'' primitives in these libraries such as thresholding, Boolean logic, discrete indexing, and sorting operations yield zero or undefined gradients that are not useful for optimization. While numerous ''soft'' relaxations have been proposed that provide informative gradients, the respective implementations are fragmented across projects, making them difficult to combine and compare. This work introduces **SoftJAX** and **SoftTorch**, open-source, feature-complete libraries for *soft differentiable programming*. These libraries provide a variety of soft functions as drop-in replacements for their hard JAX and PyTorch counterparts. This includes (i) elementwise operators such as *clip* or *abs*, (ii) utility methods for manipulating Booleans and indices via fuzzy logic, (iii) axiswise operators such as *sort* or *rank* -- based on optimal transport or permutahedron projections, and (iv) offer full support for straight-through gradient estimation. Overall, SoftJAX and SoftTorch make the toolbox of soft relaxations easily accessible to differentiable programming, as demonstrated through benchmarking and a practical case study.

General Machine Learning · Evaluation

Zhengshuyuan Tian, Chuanxin Lan, Chenxi Wang, Lei Wang, Guoxin Kang, Zhengxin Yang, Yunyou Huang, Xuehai Hong, Wanling Gao, Jianfeng Zhan

Current LLM evaluations often conflate benchmark performance with intrinsic model capability. This is misleading, as observed outcomes arise from the entire evaluation system, including datasets, prompting methods, decoding parameters, and the software–hardware stack, rather than the model alone. When this system is underspecified, attribution becomes unreliable; in practice, evaluation choices alone can induce accuracy swings of up to 70\%. This attribution challenge is compounded by the open-ended nature of LLM evaluation. Questions span languages, domains, and usage styles, forming highly variable and implicitly shifting datasets. Consequently, strong performance on static benchmarks may reflect alignment with surface patterns rather than robust underlying capability. Prior studies either focus on individual components, overlooking their interactions, or investigate manually curated and small-scale question variants, lacking a holistic perspective, precluding precise attribution of intrinsic model capabilities amidst the confounding influences. To address these limitations, we propose LLM evaluatology, a principled framework that grounds LLM evaluation in a causally informed system design. By jointly modeling evaluation components and structured question variations, it enables interpretable, reproducible, and causally faithful assessment of model capability, establishing clear conditions under which evaluation results are meaningful and trustworthy.

Optimization · Discrete and Combinatorial Optimization

Esha Singh, Dongxia Wu, Chien-Yi Yang, Tajana Rosing, Rose Yu, Yian Ma

Multi-objective combinatorial optimization seeks Pareto-optimal solutions over exponentially large discrete spaces, yet existing methods sacrifice generality, scalability, or theoretical guarantees. We reformulate it as an online learning problem over a decomposed decision space, solving position-wise bandit subproblems via adaptive expert-guided sequential construction. This formulation admits regret bounds of $O(d\sqrt{T \log T})$ depending on subproblem dimensionality \(d\) rather than combinatorial space size. On standard benchmarks, our method achieves 80--98\% of specialized solvers performance while achieving two to three orders of magnitude improvement in sample and computational efficiency over Bayesian optimization methods. On real-world hardware-software co-design for AI accelerators with expensive simulations, we outperform competing methods under fixed evaluation budgets. The advantage grows with problem scale and objective count, establishing bandit optimization over decomposed decision spaces as a principled alternative to surrogate modeling or offline training for multi-objective optimization.

General Machine Learning · Hardware and Software

Anselm Paulus, Andreas René Geist, Vit Musil, Sebastian Hoffmann, Georg Martius

Automatic differentiation (AD) frameworks such as JAX and PyTorch have enabled gradient-based optimization for a wide range of scientific fields. Yet, many ''hard'' primitives in these libraries such as thresholding, Boolean logic, discrete indexing, and sorting operations yield zero or undefined gradients that are not useful for optimization. While numerous ''soft'' relaxations have been proposed that provide informative gradients, the respective implementations are fragmented across projects, making them difficult to combine and compare. This work introduces **SoftJAX** and **SoftTorch**, open-source, feature-complete libraries for *soft differentiable programming*. These libraries provide a variety of soft functions as drop-in replacements for their hard JAX and PyTorch counterparts. This includes (i) elementwise operators such as *clip* or *abs*, (ii) utility methods for manipulating Booleans and indices via fuzzy logic, (iii) axiswise operators such as *sort* or *rank* -- based on optimal transport or permutahedron projections, and (iv) offer full support for straight-through gradient estimation. Overall, SoftJAX and SoftTorch make the toolbox of soft relaxations easily accessible to differentiable programming, as demonstrated through benchmarking and a practical case study.

General Machine Learning · Evaluation

Weijia Li, KE GAO, Jiajie Li, Han Sun, Yuhe Ding, Yongdong Mai, Yiran Le, Yongjie Qian, Zhibin Zhang, Xinyu Wang 等

Cross-architecture migration of high-performance libraries dictates ecosystem readiness on emerging hardware. The challenge is twofold: disentangling library-scale dependencies and performance-critical kernels with ISA-specific SIMD intrinsics, often trading migration speed for peak performance. While LLM-based agents offer a promising approach, are confined to function-level tasks or scalar code, failing to assess agents’ capabilities and limitations in realistic, library-scale migration. We present CLAM-Bench (Cross-architecture Library-scale Agent Migration benchmark), featuring 85 critical kernels from widely used libraries, including OpenCV, libjpeg, and NCNN. It supports comprehensive evaluations of compilability, correctness, and performance across major transitions: ARM→RISC-V, x86→ARM, and ARM→LoongArch. Evaluation of 12 SOTA agent-LLM combinations on CLAM-Bench reveals that, due to the lack of library-level navigation and hardware-aware optimization, agents regress to superficial pattern matching, yielding only 20.88% correctness and 0.83x speedup for libjpeg. Motivated by these findings, we further propose FSCM, a multi-agent framework incorporating hardware-aware global reconfiguration and performance optimization. FSCM improves OpenCV correctness to 71%. The benchmark and code are available at https://anonymous.4open.science/r/clam_bench-D8EB/.

General Machine Learning · Hardware and Software

Zifan He, Rui Ma, Yizhou Sun, Jason Cong

Modern large language model (LLM) serving increasingly depends on efficient long-context processing and generation mechanisms, including sparse attention, retrieval-augmented generation (RAG), and compressed contextual memory, to support complex reasoning. We show that these optimizations can be unified into a four-step memory processing pipeline: *Prepare Memory*, *Compute Relevancy*, *Retrieval*, and *Apply to Inference*. Through systematic profiling, we identify a 22\%-97\% memory processing overhead in LLM inference and strong heterogeneity in its computational characteristics. Motivated by this insight, we argue that \textbf{heterogeneous systems} are well-suited to accelerate memory processing and thus end-to-end inference. We demonstrate this approach on a GPU-FPGA system by offloading sparse, irregular, and memory-bounded operations to FPGAs while retaining compute-intensive operations on GPUs. Evaluated on an AMD MI210 GPU and an Alveo U55C FPGA, our system achieves $1.04\sim2.2\times$ speedup and $1.11\sim4.7\times$ energy reduction across multiple LLM inference optimizations than the GPU baseline (similar results hold on NVIDIA A100). These results establish heterogeneous systems as a practical direction for efficient LLM memory processing and inform future heterogeneous hardware design.