跳到正文
研踪PaperTrail

论文库

检索计算机领域论文,按研究方向、作者或关键词筛选。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+197新增论文
+151其中 AI 相关
+16已有论文修订
+91新增 AI 导读

arXiv 最近完成扫描:2026-11-22 08:44 · 全站最近入库:2026-11-21 12:19各信源进度 →

统计口径与数据新鲜度

统计 2026-11-21 09:35 — 2026-11-22 09:35(北京时间) · 点击数字查看对应简报;AI 相关包含在新增论文内。

最近完成检查:2026-11-22 08:44最近发现新增版本:2026-11-21 12:19库内论文最新更新时间:2026-11-20 02:59

每分钟刷新已保存的数据。空检查不代表采集异常;新收录可能包含历史论文。AI 导读按新增版本计数。

29,158 已收录论文24,181 AI 相关8,992 有修订112 HF 社区发现

浏览论文库

时间范围跟随排序:版本更新、本站收录或首次发表。最新收录可能包含历史论文。部分信源异常,查看状态 →

找到 275 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-22 08:42(北京时间)
版本更新 · v2arXiv:2606.29129

Improved Scaling for Fast Mode of Ozaki Scheme II

Shota Kawakami, Daisuke Takahashi

原文摘要

Ozaki scheme II emulates high-precision matrix multiplication using low-precision integer matrix operations based on the Chinese remainder theorem (CRT). It first scales the high-precision matrices to convert them into integer matrices. For this scaling step, Ozaki scheme II provides two modes: accurate mode, which uses INT8 matrix multiplication to estimate scaling factors, and fast mode, which applies the Cauchy--Schwarz inequality at lower computational cost. We show that the existing formula lacks scale invariance; multiplying the input matrices by a constant changes the effective bit width of the integer matrices in the scaling step, causing accuracy degradation or CRT recovery failure. To address this, we propose a revised scaling formula derived from the CRT uniqueness condition via the Cauchy--Schwarz inequality. The proposed formula is scale-invariant by construction, guarantees that the CRT uniqueness condition is always satisfied, and introduces no additional overhead over the original fast mode. Experiments on an NVIDIA GH200 GPU show that the proposed method achieves accuracy comparable to that of accurate mode while maintaining throughput comparable to that of fast mode. In the accuracy--throughput trade-off, the proposed method overcomes the accuracy limitation of fast mode and the throughput constraint of accurate mode, offering a superior accuracy and performance.

cs.MS · cs.DC · math.NA

版本更新 · v4arXiv:2512.00053

Ten-Four: An Open-Source Fused Dot Product Unit for Mixed-Precision GPGPU Tensor Cores

Nikhil Rout, Blaise Tine

原文摘要

Efficient mixed-precision MMA operations are critical for accelerating deep learning workloads on GPGPUs. However, existing open-source GPGPU Tensor Cores rely on discrete arithmetic units causing high latency, accumulated rounding errors, and poor resource utilization. To address these challenges, we propose Ten-Four, a configurable mixed-precision fused dot product (FEDP) unit microarchitecture implemented as part of the open-source RISC-V Vortex GPGPU's Tensor Core Unit extension. It supports low-precision multiplication in TF32, FP16, BF16, FP8, BF8, INT8 and INT4 with FP32/INT32 accumulation, native OCP Microscaling (MX), and sparse lane clock-gating for dynamic power reduction, while matching NVIDIA Tensor Core numerical accuracy. Ten-Four achieves 4-cycle latency at 300 MHz on the AMD Alveo U55C FPGA, delivering 2.7$\times$--7.9$\times$ speedup over Berkeley HardFloat and FPnew at less than 60% the area cost. We further demonstrate its format extensibility by prototyping LNSF4, an NVFP4-variant matching its perplexity with 6.7% fewer LUTs and 11.1% lower dynamic power.

cs.AR

新论文 · v1arXiv:2611.11605

ScanDreamer: A Map-Grounded World Model for Quadrotor Navigation

Yunyi Gao, Yixing Shi, Yize Liu, Hai-Duong Tran, Nhat-Hoang P. Nguyen, Shuai Wang, Yi Deng, Shengming Liu 等 11 位作者

原文摘要

Quadrotors inspecting buildings and tunnels must navigate unknown clutter with a forward-facing depth camera whose limited field of view constrains scene understanding and collision-avoiding navigation. Recent world-model agents learn navigation through imagined experience (e.g., AirDreamer), but remain data-intensive and often lack explicit geometric memory and awareness of what the camera can observe. We propose ScanDreamer, a map-grounded world model that constructs an online 2D scan-memory map from depth images and vehicle pose. Our world model, actor, and critic are conditioned on this map, which also defines a geodesic progress reward. Moreover, a forward-cone penalty discourages horizontal motion outside the camera's viewing direction. During imagination, ScanDreamer propagates the quadrotor pose over a frozen map and derives progress and wall-contact signals from the mapped geometry. We train ScanDreamer for 0.8 million environment steps (5.5 h on one GPU) and evaluate it in NVIDIA Isaac Sim over 1560 episodes spanning unseen layouts and parameter ranges. Across three seeds (mean $\pm$ SD), ScanDreamer achieves 96.3 $\pm$ 0.3% success with 3.6 $\pm$ 0.3% collision, compared with 59.6 $\pm$ 3.0% success and 28.8 $\pm$ 2.8% collision for AirDreamer at the same online-interaction budget. Removing the forward-cone penalty increases collisions to 9.9 $\pm$ 3.1%. These results show that combining explicit map memory, map-grounded imagination, and perception-aware motion substantially improves navigation performance and data efficiency. The source code is available at https://github.com/WANGSHUAIzzzz/ScanDreamer.

cs.RO

版本更新 · v2arXiv:2610.33932

FastSAR: Open-Source SAR Image Formation on CPUs, GPUs, and TPUs, and a Matched Comparison of Algorithms, Precisions, and Processors

Paul Singerman, Tom Braun

原文摘要

The number of spaceborne synthetic aperture radar (SAR) collectors is growing exponentially, each returning raw collections that must be formed into images. Exact backprojection needs 1.6 trillion interpolated reads and 90 trillion floating-point operations for a 108-megapixel Umbra image. How long it takes to form an image, and at what price, depends on the algorithm, the arithmetic, and the hardware. Open-source implementations have never been compared on matched hardware against a common reference, and none run on Google tensor processing units (TPUs). This paper presents FastSAR, an open-source SAR image formation library for central processing units (CPUs), NVIDIA graphics processing units (GPUs), and TPUs. With it, we compare three SAR image formation algorithms, four floating-point formats, and six cloud devices in every supported combination, with cost per image at on-demand cloud prices as the fair common measure. FastSAR is 1.06 to 22,000 times faster than the tested open-source implementations at equal or better accuracy, and the full suite of hardware, algorithm, and floating-point combinations lets a user choose the Pareto-optimal configuration for their balance of accuracy, speed, and cost. The L4 GPU forms images the cheapest overall. Larger GPUs are faster but cost more per image because their price rises faster than their speed. The TPUs come close to the GPUs in cost, but only for the factorized algorithm, and the CPU is far less efficient in both speed and cost. Lower precision formats buy less of a speedup than expected, and each step down in precision costs accuracy.

cs.PF · eess.SP

新论文 · v1arXiv:2611.13004

G-STAR: Group-Scaled Token Capacity for Deeply Compressed Autoregressive Generation

Xu Huang, Ye Huang, Xiaotong Li, Zhengrong Yue, Jialv Zou, Minhao Jing, Haodong Li, Wenhao Yan 等 11 位作者

原文摘要

Discrete autoregressive models offer a scalable framework for image generation. However, they face a fundamental trade-off: long token sequences preserve representation quality but slow generation, while aggressive spatial compression reduces representation capacity. Based on product quantization, we show that, under a fixed nominal information budget, group size should scale quadratically with the per-axis spatial compression factor. Building on this principle, we propose GSTAR, a spatial--group co-scaling framework that enables \(32\times\) and even effective \(64\times\) compression while preserving strong reconstruction, understanding, and generation performance. Since larger groups introduce intra-group dependencies that must be modeled efficiently, we analyze their impact on inference latency and generation quality and introduce a semi-autoregressive decoder that combines parallel feature computation with lightweight causal modeling. With these designs, our highly compressed models perform competitively with leading models across five representative text-to-image benchmarks (GenEval, DPG-Bench, PRISM, LongText-Bench, and Qwen-Image-Bench). At \(1024\times1024\) resolution, our model delivers a \(20\times\) speedup over prior discrete autoregressive models, generating each image in 2.5 seconds on a single NVIDIA H200. Together, these results show that highly compressed discrete representations can enable high-quality autoregressive generation at low latency.

cs.CV

新论文 · v1arXiv:2611.11426

Vietnamese Legal Ontology: Every legal answer, traced to the article that grounds it

Khac-Hoai Nam Bui, Quan Tran Minh

原文摘要

A legal answer is only as good as the article it rests on. Vietnam's largest open legal corpus pairs 109,543 published questions and answers with 573,851 legal documents, yet its citations are free text: "Khoan 3 Dieu 101 cua Bo luat nay" ("Clause 3, Article 101 of this Code") names neither the code nor its version. Following the five stages of NVIDIA Auto Ontology, Discover, Ground, Validate, Expose and Learn, we turn four open datasets into a knowledge graph of 11.5 M nodes and 16.0 M edges in which every node and edge carries its source record and method, with no duplicate keys, dangling edges or schema violations. Of 67,868 cited provisions, 61,487 resolve to a dated law version, 51,694 to a parsed article, clause or point, and 19,012 are confirmed word for word. The ontology then shows what flat records hide. 36% of the 78,361 answers whose citations resolve now cite a law that is no longer in force; 2,525 verified citations point to an article number that says something else in the current law; 2,520 citations were re-versioned because their words match a different version than their date suggests; and 5,575 document numbers are shared by several local authorities, one of them by 72. Every number in this paper is computed from the graph.

cs.CL · cs.AI

版本更新 · v2arXiv:2604.24549

GradMAP: Gradient-Based Multi-Agent Proximal Learning for Grid-Edge Flexibility

Yihong Zhou, Hongtai Zeng, Thomas Morstyn

原文摘要

Coordinating large populations of grid-edge devices requires learning methods that remain decentralised in deployment while still respecting three-phase AC distribution-network constraints. This paper proposes gradient-based multi-agent proximal learning (GradMAP) to address this challenge. GradMAP trains independent neural network policies for each agent that makes online decisions without communication. During offline training, GradMAP embeds a differentiable power grid environment and uses implicit differentiation to propagate exact AC power flow gradients. To speed up training while maintaining performance, GradMAP reuses expensive environment gradients through a proximal surrogate within a trust region defined in the more direct policy-output space, instead of the probability distribution space used in PPO. Using a single workstation-class NVIDIA RTX PRO 5000 Blackwell GPU with 48~GB of memory, GradMAP trains 1,000 agents on the IEEE123 network in 15 minutes and 100,000 agents on the IEEE8500 network in 40 minutes, without parameter sharing. In both cases, GradMAP achieve the lowest mean daily maximum voltage violation and the lowest energy cost among all evaluated methods, including standard multi-agent reinforcement learning methods and gradient-based learning methods extended to the multi-agent setting by us. The training time to convergence is also the lowest, with a 5x to more than 10x training speedup compared to other methods.

cs.LG · cs.AI

新论文 · v1arXiv:2611.11291

Multi-Node Transformers for Object Detection and Captioning Across Distributed Edge Networks

Saurabh Kumar, Dewant Katare, Prakriti Jain

原文摘要

This paper presents a unified transformer-based framework for object detection and dense region captioning across distributed edge networks. The architecture uses a shared visual encoder followed by detection, region extraction, and autoregressive language-decoding stages, allowing localization and region description to reuse the same image features. This design removes the second image-encoding path required by decoupled detection and captioning pipelines. To support execution on resource-limited devices, a workload-allocation formulation assigns model stages and input streams across connected edge nodes while accounting for execution time, memory capacity, processor utilization, power limits, and intermediate tensor transfer cost. The framework is evaluated on a six-device edge testbed comprising four NVIDIA Jetson Orin Nano and Orin NX inference nodes and two Raspberry Pi 5 auxiliary CPU nodes, with TensorRT FP16 and INT8 execution on the NVIDIA devices. On a Jetson Orin Nano, the unified INT8 model achieves 38.1 FPS with 3.5 GB peak memory use and 11.8 W measured power. Under four concurrent streams, the proposed allocation achieves 35.2 FPS and 28.4 ms mean end-to-end latency with an internode transfer rate of 6.8 MB/s.

cs.DC · cs.CV · cs.LG

新论文 · v1arXiv:2611.11289

A Unified Framework for License Plate Identification and Character-Recognition

Saurabh Kumar, Dewant Katare, Prakriti Jain

原文摘要

This paper presents a unified framework for license plate detection and character recognition on low-power edge hardware. The proposed network uses a shared CSP-Darknet and PANet feature encoder with an anchor-free localization head and a RoIAlign-based sequence-recognition branch. Character decoding is performed directly on aligned feature maps, avoiding image-space plate cropping and a second recognition backbone. The model is trained on an AI-assisted dataset of approximately 37,000 manually verified images covering license plate layouts from all 50 United States. The dataset includes variations in illumination, viewpoint, weather, blur, occlusion, and plate design. The framework is converted to an INT8 TensorRT engine and evaluated on an NVIDIA Jetson Orin Nano with 8 GB memory. The INT8 model shows an mAP50 of 97.8% and a full-sequence accuracy of 90.7%. It records 147 FPS at 6.8 ms end-to-end latency, with 315 MB inference memory and 9.4 W average power consumption. Relative to the YOLOv8s+CRNN baseline, the proposed deployment provides a 7.1x throughput increase and a 77.8% reduction in allocated inference memory

cs.LG · cs.AI · cs.CV

新论文 · v1arXiv:2611.11200

Source-Anchored Visual Autoregression for Extreme Image Compression

Bingxin Wang, Yuan Lan, Yang Xiang, Yiqing Huang

原文摘要

In the extreme-low bitrate range below $0.01$ bpp, image compression becomes a constrained generation problem: the reconstruction depends on inferring plausible missing details while staying close to the input image. Autoregressive token models provide an efficient coarse-to-fine prior for this setting, but the transmitted early-scale tokens often leave object identity and texture underdetermined, which can lead to source-inconsistent reconstructions. We introduce Source-Anchored Visual Autoregression (SAVAR), a generative codec built on Visual Autoregressive (VAR) modeling. SAVAR transmits early-scale tokens and a compact caption. The tokens provide image-specific layout information, while the caption supplies global semantics. SAVAR reconstructs a partial VQ-VAE latent from the retained scales, maps it to the start-of-sequence initialization, and uses the same latent to condition the caption embedding before adaptive layer normalization, separating two roles that are typically driven by a shared condition in VAR-style generation. On the Kodak and CLIC2020 datasets in the extreme-low range, SAVAR reduces perceptual distortion relative to matched ARPC retained-scale operating points, lowering LPIPS by about $8$--$9\%$ and DISTS by up to $22\%$ at a lower bpp. It reconstructs a $512 \times 512$ image in about $218$ ms on one NVIDIA L20X GPU, retaining fast autoregressive decoding in the evaluated setting.

cs.CV

新论文 · v1arXiv:2611.11169

Map2SLAM: From Streaming Map Prediction to Real-Time SLAM

Jian Gao, Lin-Zhuo Chen, Yihang Chen, Jiagang Chen, Jiacheng Dong, Xun Cao, Nan Xue, Yinghao Xu 等 9 位作者

原文摘要

Streaming 3D foundation models (3DFMs) estimate camera poses and dense geometry online, but they fall short of SLAM: their predictions are fixed once emitted, so loop closures cannot correct accumulated drift. With such models already providing strong geometric capability, a central question for SLAM becomes how to schedule their inference and how to keep its results as revisable state. We present Map2SLAM, which keeps LingBot-Map frozen, trains only a lightweight retrieval head, and builds SLAM on the model's inference interface, turning its predictions into observations of an external persistent SLAM state. The front-end tracks every incoming frame with the model's causal streaming inference and decides which frames are written to the persistent state as keyframes. The back-end asynchronously decides when and which past frames to re-observe, re-running the model's joint inference over them with fresh context; verified loop observations correct the trajectory online, and segment observations refine dense geometry after the sequence. All invocations share a single frame encoder pass per image, whose cached features also drive loop retrieval, so frequent re-invocation incurs no additional encoding cost. Map2SLAM achieves state-of-the-art performance on KITTI, VBR, and Oxford Spires, reducing mean ATE by over 40% compared with the best uncalibrated baselines, while maintaining real-time tracking at over 28 FPS at a resolution of $518\times154$ on an NVIDIA RTX 5090D GPU. Please visit our project page at https://robbyant-research.github.io/Map2SLAM/.

cs.CV

版本更新 · v2arXiv:2610.36558

Holomorphic Cauchy Networks for Efficient Stationary Tail Computation in High-Dimensional Reflected Brownian Motions

Kehan Wang

原文摘要

Stationary tail computation for high-dimensional reflected Brownian motions (RBMs) remains computationally challenging. Building on NN4MGF's shared-coordinate basic-adjoint-relationship (BAR) framework, we develop Holomorphic Cauchy Networks (HCNs). Native complex-valued rational Cauchy subnetworks enforce holomorphicity on pole-free domains, removing sampled Cauchy-Riemann penalties and their input-derivative graphs. Exact origin anchoring and boundary-coordinate exclusion enforce structural identities, while anchored interactions represent dependent transforms. Under holomorphic-log assumptions, we establish simultaneous value and first-derivative approximation for the constrained architectures and a mixed-derivative lower bound for coordinate-only representations under nonzero cross-covariance. Closed-form real-axis derivatives and sequential backpropagation further control memory. On the same NVIDIA H100 NVL, the tested 30D HCN configuration uses approximately tenfold less peak allocated memory and has 9.7-fold lower mean update time than NN4MGF; extrapolation to the respective scheduled update counts gives a projected full-schedule update-time ratio of 58.2. With 64 million sampled base inputs per high-dimensional run, HCN reaches 50 dimensions on a laptop GPU with 6.367% mean relative error over thresholds 1 through 8. Degree-matched comparisons yield lower reported mean relative errors over the primary evaluation ranges, including 1.422% versus 6.684% on the 2D Harrison non-product benchmark. On a 10D correlated RBM benchmark, HCN achieves 3.829% mean relative error over thresholds 1 through 8 against a simulation reference, compared with 19.386% for an independent-marginal baseline.

cs.LG · math.NA · math.PR

新论文 · v1arXiv:2611.11023

Where Should Models Run? A Collaborative Inference Framework for Embodied Intelligence in 6G

Wanlu Sun, Zhenhui Yuan, Xiang Ji, Yi Han, Qing Li, Gan Zheng

原文摘要

Embodied intelligence systems rely on fast and reliable inference to enable real-time perception, reasoning, and control. Recent advances in large-scale artificial intelligence (AI) models, including Large Language Models (LLMs) and Vision-Language-Action (VLA) models, have significantly enhanced the capabilities of embodied agents in complex environments. However, deploying these models remains challenging due to the limited computational resources and energy budgets of robotic platforms. A fundamental question is where inference should be executed. Collaborative inference, which partitions models across end devices, edge servers, and the cloud, has emerged as a promising solution. Yet determining the optimal partitioning point remains an open problem, as it depends on wireless network conditions, model structure, and privacy requirements. This article reviews the state of the art in collaborative inference for embodied intelligence and proposes an edge-centric framework in which the edge coordinates model partitioning and execution. Through both simulation and real-world experiments over WiFi and 5G networks, we demonstrate that the optimal partitioning strategy varies with network conditions and model structure. We further identify latency-minimizing partition points for ResNet-101, Qwen2.5-VL-7B, and OpenVLA-7B in collaborative inference between NVIDIA Jetson Nano or AGX Xavier devices and an AGX Thor edge server under varying wireless conditions. Finally, we discuss the future directions for collaborative inference in 6G-empowered embodied intelligence systems.

cs.LG · cs.AI

新论文 · v1arXiv:2611.10984

Network Contention State Aware Dynamic Optimizer for Accelerator Collective Communication Libraries

Rajat Bisht, Tevfik Kosar, Anthony Skjellum, Gene Cooperman, MD S Q Zulkar Nine

原文摘要

Collective Communication Libraries (CCL) are critical for large-scale distributed AI/ML training, yet their configurations are determined by static heuristics that do not account for real-time network conditions. Our empirical study shows that these heuristics perform near-optimally in contention-free environments, but degrade substantially under background traffic. To address this, we present a cooperative multi-agent run-time optimizer for accelerator collective communication. Our approach employs a two-agent architecture: (i) detector- a Long Short Term Memory(LSTM) recurring neural network fed classifier head with a sequential probability ratio test that distinguishes sustained contention from transient spikes; and (ii) selector- a configuration selector built on a Dueling Double DQN (D3QN) with Conservative Q-Learning that adapts configuration parameters in response. We implemented this dynamic tuner over NCCL, a widely used CCL developed by NVIDIA. When evaluated on Chameleon Cloud's isolated bare-metal cluster, this dynamic tuning approach reduces AllReduce latency by 36% over NCCL and 32% over a fully converged AutoCCL. We further evaluated it across six real AI workloads spanning convolutional networks, transformers, and recommendation architectures. Our multi-agent tuner approach delivers up to 65% speedups on Delta cluster with NVIDIA A100 GPUs and Slingshot-11 fabric internode connectivity.

cs.DC · cs.LG · cs.MA

新论文 · v1arXiv:2611.10688

Performance Benchmarking of Federated Learning Frameworks for Medical Image Classification

Vithor Pileco, Luan Mantegazine, Claudio Geyer

原文摘要

Selecting a Federated Learning (FL) framework requires understanding the resource demands of its training and evaluation workflow. This work compares Flower, NVIDIA FLARE, and FedBioMed using two-dimensional MedNIST classification with ResNet-10 in a containerized, single-host environment. Following a dated review of repository activity and documented medical applications, we evaluated nine configurations per framework, with three random seeds, for a total of 81 executions. The configurations varied federation size, communication rounds, and batch size. We monitored execution time, CPU and RAM consumption, host-level GPU utilization, server network traffic rates, and model accuracy at the end of each execution. The implementations shared nominal learning settings while retaining differences in data splitting and federated evaluation. Configuration-blocked non-parametric analysis identified differences in execution time, client and server memory, GPU utilization, and network traffic rates. NVIDIA FLARE exhibited the lowest server-memory consumption and the highest host-level GPU utilization, Flower exhibited the highest average server network traffic rate, and FedBioMed required more client memory and longer execution time. No statistically significant difference in final accuracy was detected within the evaluated short training schedules; this does not establish convergence or equivalent predictive performance. The results characterize the configured workflows under shared hardware resources. They provide a concrete case study of system behavior, with conclusions bounded by workflow asymmetries, local networking, and the lightweight MedNIST workload.

cs.DC · cs.PF

版本更新 · v4arXiv:2608.10317

From Detection to Understanding: TAR and TAR-Bench for Multi-Task Traffic Anomaly Reasoning

Han Zhang, Yilin Zhao, Zaid Pervaiz Bhat, Zheng Tang, Varun Praveen, Vidya N. Murali, David C. Anastasiu, Tomasz Kornuta

原文摘要

Detecting a traffic anomaly does not establish whether a video-language model can explain what happened, localize it in time, or identify its causes. We introduce TAR (Traffic Anomaly Reasoning) and TAR-Bench, paired resources for training and evaluating these complementary capabilities across 10 tasks spanning question answering, temporal reasoning, and scene understanding. TAR contains 44,040 automatically generated annotations with chain-of-thought traces for 3,670 CCTV videos from eight public datasets. TAR-Bench provides 960 human-curated annotations for 80 held-out clips from 17 public YouTube videos. Evaluation of eleven vision-language models and analysis of fine-tuned outputs reveal temporal, causal, and descriptive errors not captured by closed-form question-answering accuracy. Training on all 10 tasks raises mean benchmark scores from 34.3 to 55.7 for Cosmos-Reason2-8B and from 30.9 to 53.9 for Qwen3-VL-8B-Instruct. These results support joint training across complementary tasks, while highlighting persistent limitations in temporal precision and causal attribution. TAR and TAR-Bench serve as the official training and in-domain evaluation resources for AI City Challenge 2026 Track~3. The dataset is available under CC-BY-4.0 at https://huggingface.co/datasets/nvidia/PhysicalAI-Traffic-Anomaly-Reasoning.

cs.CV

新论文 · v1arXiv:2611.10469

Hibana: Fast Agentic Inference on Unified-Memory Devices

Yahya Emara, Chi-Chih Chang, Anthony Fei, Ahmed F. AbouElhamayed, Chong-Yan Chen, Hung-Yueh Chiang, Kai-Chiang Wu, Mohamed S. Abdelfattah

原文摘要

Local AI promises capable personal assistants that run on users' own devices, keeping data and computation under their control. Open-weight models that fit on local devices can now handle practical tasks such as coding. Yet current inference engines fall short. Popular engines such as vLLM and SGLang target datacenter GPUs with dedicated host memory, while local engines underuse the memory hierarchy of unified-memory devices. We present Hibana, an inference engine for unified-memory devices that co-optimizes model representation, memory management, and decoding strategy. Model surgery, the process of editing a trained model's architecture to suit its hardware, changes how the model computes, shrinking its memory footprint and data movement. Unified-memory management preserves each agent's state across turns, so long-running agents can resume instantly instead of recomputing their state. Adaptive speculation generates several tokens per step. Hibana codesigns these features around the shared memory capacity and limited bandwidth of unified-memory systems-on-chip to support local multi-agent workloads. On a single DGX Spark serving Qwen and Nemotron models of up to 125B parameters, Hibana completes up to 3.4$\times$ more agentic tasks per hour and delivers up to 3.9$\times$ higher decoding throughput than other established frameworks such as vLLM, NVIDIA Flare, llama.cpp, and TensorFold. Code is available at https://github.com/sharar-project/hibana.

cs.CL · cs.AI

新论文 · v1arXiv:2611.09825

BitDecoding-FP4:通过算法-系统协同设计的原生FP4 KV缓存解码

BitDecoding-FP4: Native FP4 KV Cache Decoding via Algorithm-System Co-Design

Dayou Du, Woon Yee Ng, Guangming Sheng, Nick Lau, Jianyi Cheng, Luo Mai

AI 速读 基于摘要

提出算法与系统协同设计的原生FP4 KV缓存解码方案,在保持接近BF16精度的同时提升长上下文LLM服务吞吐。

cs.AI · cs.AR · cs.CL · cs.PF

查看原文摘要

The key-value (KV) cache dominates memory capacity and bandwidth in long-context LLM serving. While 4-bit microscaling formats (NVFP4, MXFP4) shrink the cache and run attention on FP4 Tensor Cores, a fundamental gap remains between algorithmic quantization and system-aware acceleration: existing FP4 KV systems use vanilla quantization that causes severe accuracy drops, whereas accuracy-oriented 4-bit algorithms rely on vector codebooks whose lookup overhead degrades decoding throughput. We present BitDecoding-FP4, an algorithm--system co-design for accurate and efficient native FP4 KV decoding. We mitigate three error sources: Dual-Residual Query Quantization splits queries into two FP4 tensors, cutting query logit error from 60% to 1%; Underflow-Aware Probability Scaling prevents E4M3 scales from rounding long-context probabilities to zero; and Outlier-Aware Key Quantization combines static smoothing, Hadamard rotation, and scale search to suppress key outliers under token-wise scaling. At the system level, we remove the softmax bottleneck via transposed softmax with fused probability quantization and stacked query MMA on paged KV caches. Across dense and MoE models with GQA and MLA, BitDecoding-FP4 stays within 0.5 points of BF16 KV accuracy with NVFP4 (1 point with MXFP4), and tracks BF16 KV in RL training rollouts. In end-to-end serving on NVIDIA B300 with Qwen3-4B and Qwen3-30B-A3B, it delivers 2.4-3.0$\times$ higher peak serving throughput than BF16 KV. The implementation is available at https://github.com/OpenBitSys/BitDecoding-FP4.

版本更新 · v2arXiv:2606.17241

超越基准:嵌入式车载硬件上交通标志识别的连续边缘推理

Beyond Benchmarks: Continuous Edge Inference for Traffic Sign Recognition on Embedded Vehicular Hardware

Aditya Mishra, Anushka Tailor, Haroon Lone

AI 速读 基于摘要

作者提出面向NVIDIA Jetson Orin Nano的连续边缘推理系统Edge-TSR,并在流式视频评测中揭示图像基准与真实部署之间的性能落差。

cs.CV · cs.RO · eess.SY

查看原文摘要

Traffic sign recognition (TSR) systems deployed on embedded vehicular hardware face deployment effects that image-based benchmark evaluation does not capture, including frame-to-frame prediction instability in streaming video and load-dependent throughput under sustained operation. We present Edge-TSR, a continuous edge inference system for TSR on the NVIDIA Jetson Orin Nano that combines sparse detection, multi-object tracking, region-level fine-grained classification, and a lightweight track-aware temporal stabilization mechanism adding only 0.07~ms per sampled frame. Evaluating on the source videos of the INTSD benchmark, we find that the state-of-the-art TSR models lose 20-30 percentage points of mAP@50 when moving from image-based to streaming evaluation. Edge-TSR achieves the best streaming classification performance among all evaluated baselines, and temporal stabilization improves accuracy by 10.58 percentage points over per-frame inference. We characterize latency, throughput, and thermal behavior across sampling intervals and four driving conditions, showing that throughput is load-proportional (20 ms per classified region) and that every-frame inference is limited by throughput rather than thermal throttling. A 55-minute, 26 km vehicular deployment demonstrates sustained online operation at 16.18 FPS within safe thermal limits without cloud offload. We also release the annotated 50,732-frame streaming evaluation dataset and full implementation.

新论文 · v1arXiv:2611.09651

Concurrent Use of FP32 and FP64 Computing Units on GPUs and Mixed-Precision Placement in WENO Reconstruction

Jun Hatayama, Daichi Mukunoki, Yusuke Onishi, Atsushi Sekimoto, Toshinori Kouchi

原文摘要

Scientific computing often requires FP64 arithmetic, although FP64 performance is limited on NVIDIA GPUs. Double-float (DF) emulation, which represents a value using two FP32 values with about 48 bits of significand, outperforms native FP64 where FP64 units are limited. On the other hand, mixed-precision strategies offload to FP32 the operations that do not determine accuracy. In both approaches, the native FP64 units remain idle during FP32 instruction execution. We investigated whether this idle resource can be recovered on A100, GH200, GB200, and RTX PRO 6000 Blackwell Max-Q. A microbenchmark shows that co-resident FP32 and native-FP64 kernels exceed the throughput of the faster kernel alone on A100 but not on GH200 and GB200; on RTX 6000 Blackwell, they do so against FFMA but, against DF emulation, only when the issue density of the DF kernel is reduced. We then tested whether this gain carries over to computing the WENO-Z reconstruction with the nonlinear weights in FP32 and the candidate polynomials in native FP64 (W32P64). We observe substantial overlap on A100, whereas GH200, GB200, and RTX 6000 Blackwell show little overlap. Importantly, the fastest placement does not always preserve the design accuracy of WENO-Z: for a one-dimensional entropy wave problem, W32P64 fails to achieve the design order of accuracy for WENO7 and WENO9. In contrast, DF emulation of the entire reconstruction comes close to it on the grids tested for WENO5, WENO7, and WENO9 while delivering a $2.51-3.94 \times$ speedup over FP64 on RTX 6000 Blackwell. For WENO5 on the two-dimensional Euler vortex problem the W32P64 error at $n=256$ is $1.26 \times$ that of FP64 and 1/104 of that of the all-FP32 reconstruction.

cs.DC

新论文 · v1arXiv:2611.09624

TPIC: Energy-Conserving Space Plasma Simulations on GPU Supercomputers

Luca Pennati, Ivy Peng, Stefano Markidis

原文摘要

Artificial heating and numerical damping affect particle energization in kinetic space plasma simulations, even when the numerical scheme remains stable. Implicit and semi implicit energy conserving Particle-in-Cell (PIC) methods eliminate these errors in the discrete energy balance while relaxing explicit stability constraints. We present TPIC, a massively parallel implementation of the Energy Conserving Semi-Implicit Method (ECSIM) for GPU supercomputers. We extend the exact energy conservation in the ECSIM formulation to closed domains with conducting boundaries by deriving and canceling a spurious term in the discrete Poynting flux theorem. TPIC executes the complete particle and field cycle on NVIDIA and AMD GPUs, combining matrix-engine acceleration with asynchronous communication and computation. We validate TPIC in a variety of electrostatic and electrodynamic test cases relevant for space plasma physics applications, including quasineutral plasma expansion, two-stream instability, Whistler waves, Weibel instability, Whistler turbulence, and three-dimensional magnetic reconnection with conducting walls and 12.3 billion particles. We show that in all scenarios, TPIC agrees with the corresponding analytical theory or reference numerical results, and preserves the total system energy to machine precision. On 4,096 AMD GPUs, TPIC advances 1.9 trillion particles with 96.8\% weak scaling efficiency.

cs.CE · cs.DC · cs.PF

新论文 · v1arXiv:2611.09490

CucumberVision at the Edge: Accuracy, Throughput, and Energy of RGB-D Cucumber Sizing on Jetson Orin Nano and Raspberry Pi 4

Manveen Kaur, Rajmeet Singh, Saeed Mozaffari, Majid Ahmadi, Shahpour Alirezaee

原文摘要

Non-contact fruit sizing from RGB-D imagery reaches sub-centimetre accuracy on workstation hardware, but in-field grading requires the same pipeline to run on a low-power single-board computer, and the cost of that transition is rarely reported. We deploy a curvature-aware cucumber length-measurement pipeline (YOLO26n detection, prompt-based segmentation and a medial arc spline estimator) on an NVIDIA Jetson Orin Nano and a Raspberry Pi 4, and characterize both against an RTX 3050 Ti workstation by replaying 50 identical greenhouse captures of 7 specimens, with specimen-level cluster-bootstrap statistics. Accuracy is preserved: mean absolute error (MAE) is 0.58 cm (95% CI 0.46-0.69) on both the workstation and the Jetson at matched precision, with a paired difference of zero. TensorRT FP16 compilation cuts Jetson detection latency 2.57x, giving two operating points on one board: 72 ms per fruit at 0.84 cm, or 2.74 s at 0.64 cm. Four ablations map the design space. For prompted segmentation backends, measurement accuracy is fixed by the backend and reproduces across platforms to within 0.01 cm, while its cost is fixed by the platform and varies 214x, so the two choices are separable. Whole-graph INT8 quantization of the NMS-free detection head fails structurally, returning no detections, because one output tensor carries both pixel coordinates and confidences; we show that no calibration strategy can avoid this. Excluding the head restores function at 1.37x. Energy per fruit is 42.2 J on the Jetson, and its 7 W mode consumes 73% more than its 15 W mode. The Raspberry Pi with MobileSAM costs 2.5x more per fruit than the Jetson with SAM despite drawing 47% less power, and it runs without thermal throttling under sustained load. The results define an accuracy-throughput-energy envelope with three non-dominated configurations and give rules for selecting among them.

cs.RO

新论文 · v1arXiv:2611.09224

HMMF-Net: A Lightweight Hierarchical Multi-Frame Multimodal Fusion Network for Accurate Fish Feeding Intensity Assessment

Phi-Hung Hoang, Van-Manh Tran, Nam-Thuan Trinh, Thi-Thu-Hong Phan

原文摘要

Automated fish feeding intensity assessment (FFIA) is important for precision aquaculture and feed cost optimization. However, unimodal methods are vulnerable to acoustic noise and visual disturbances, while existing multimodal frameworks often rely on computationally intensive 3D convolutions or Transformers, limiting their use on edge devices. To address these limitations, we propose HMMF-Net, a lightweight multimodal architecture that combines early multi-frame visual encoding and multi-axis acoustic modeling through a hierarchical bidirectional messenger cross-modal fusion (BMCF) mechanism. The visual branch captures temporal variations in feeding behavior by stacking temporally distant frames as a multi-channel input to an adapted 2D EfficientNetB0 backbone. The acoustic branch employs adaptive multi-axis residual units (AMARU) to extract temporal, frequency, and local time-frequency representations from log-mel spectrograms. BMCF progressively exchanges information between the two modalities at intermediate representation levels through compact messenger tokens, reducing the computational overhead of dense cross-modal attention. Evaluated on the AV-FFIA dataset using five-fold cross-validation, HMMF-Net achieves 96.77% accuracy and 96.88% F1-score on the held-out test folds, with 3.987M parameters and 1.002 GFLOPs. Compared with the best-performing visual configuration (95.49% accuracy), HMMF-Net reduces the classification error by 28.38%. Deployment on an NVIDIA Jetson Orin Nano using TensorRT with FP32 precision maintains comparable accuracy, with an inference time of 11.54 ms per 2-second sample and a model size of 18.81 MB. These results demonstrate that HMMF-Net combines effective multimodal representation learning with computational efficiency, enabling accurate and practical FFIA on resource-constrained edge platforms.

cs.CV · cs.AI

新论文 · v1arXiv:2611.08928

UAV-Based Computer Vision for Cattle Detection and Feed Bunk Spatial Analysis

Pedro Oliveira Torrente, Diego Giaretta de Paulo, Lucas Pascotti Valem, Jianglong Yan, Everton Castelão Tetila, Jayme Garcia Arnal Barbedo, Liang Zhao

原文摘要

Beef cattle production is central to Brazilian agribusiness and faces growing demands for efficiency, sustainability, and animal welfare. Precision Livestock Farming (PLF) offers tools to address these challenges, particularly through monitoring feeding-related spatial patterns in confined systems. Feeding management plays a decisive role in animal performance, yet its evaluation still relies largely on subjective, labor-intensive visual inspections that are hard to scale and prone to bias. This work proposes a computer vision framework based on Unmanned Aerial Vehicles (UAVs) to quantify cattle alignment and spatial dispersion at the feed bunk under real-world feedlot conditions. A UAV dataset was collected during flight campaigns at a commercial feedlot in Mato Grosso do Sul, Brazil. Images were manually annotated into cattle head and cattle back classes to support object detection and image-level spatial analysis of cattle distribution. You Only Look Once (YOLO) models, namely YOLOv8 and YOLOv11, were evaluated, with YOLOv8 small used for the downstream spatial analysis. The feed bunk region was segmented using the Segment Anything Model (SAM), and geometric metrics based on detected head positions were developed to characterize alignment, spacing regularity, and clustering. The selected detector achieved 84.30% mAP50 and 35.50% mAP50-95, with an inference time of approximately 2.7 ms per image on an NVIDIA RTX 5060 Ti GPU. The resulting metrics distinguished representative regular and irregular spatial distributions by Root Mean Square Deviation - RMSD (1.06 m and 4.26 m), Coefficient of Variation - CV (42.70% and 96.31%), and clustering (0 and 2 clusters), respectively. These findings indicate that the proposed approach can provide indicators for assessing feed bunk spatial patterns in PLF.

cs.CV

新论文 · v1arXiv:2611.08739

Stochastic AI Model Placement for AI-and-RAN Coexistence in 6G Networks

Alaa M. Altarazi, Mohammad J. Abdel-Rahman

原文摘要

This paper considers AI-service placement over geographically distributed RAN sites equipped with GPU resources partitioned into fixed, isolated multi-instance GPU (MIG) profiles. We propose C$^2$AIR, a chance-constrained optimization model that maximizes expected AI requests served while protecting uncertain RAN workloads and enforcing probabilistic capacity reliability requirements for admitted AI services. C$^2$AIR allows a service to run at multiple sites and accounts for discrete profile, memory, and GPU-slice constraints. We evaluate C$^2$AIR on a four-site deployment calibrated from NVIDIA Aerial RAN capacity figures and MLPerf Inference results, and on seven larger scenarios with up to 15 sites and 40 AI services, computing every violation and overload probability exactly from the demand distributions. C$^2$AIR keeps the RAN violation probability within its target at every hour of the day, whereas reserving for the mean RAN demand violates it in 24\% to 47\% of the intervals. The reliability condition prevents placing AI models that would be overloaded in 85\% of the intervals at an event hour, and multi-site replication serves 2.7 times more AI requests than single-site deployment. The evaluation also shows that coarse MIG instances and a heavy demand tail can make strict RAN protection expensive.

cs.NI

新论文 · v1arXiv:2611.08720

The GPU Needs an Operating System

Reza Nourai, Behnam Bastani

原文摘要

AI chips sit mostly idle while serving models, because replies must be fast and each step waits on memory. On two large open models, the math units were busy only about 6-11% of the time. We argue the GPU needs an operating system that keeps urgent work on schedule and fills idle capacity with other useful work. A feature that ships today recovers a quarter of a GPU with work designed to share, and a tenth with real jobs, at 5% slower decode. Our research system does better: one resident kernel runs a whole model faster than vLLM's best for a single request, and shares the GPU at under 2% slowdown. Scaling beyond one hand-built model needs compilers and a few new chip features, which we specify.

cs.DC

新论文 · v1arXiv:2611.08713

ResServe: Accurate and Efficient FP4 LLM Inference with Residue Compensation

Shuo Rong, Ruibo Fan, Jiawen Wen, Xiaowen Chu

原文摘要

Block-scaled 4-bit floating-point (FP4) quantization reduces large language model inference costs but can degrade task quality. Higher-precision computation can mitigate this loss at the expense of some of the efficiency gains from FP4 inference. Narrow matrix computations common in decoding often leave spare hardware capacity for low-cost precision recovery. ResServe combines an FP4 activation base with an FP4-quantized residue representing information lost during quantization. It uses this spare capacity to compute the residue's contribution and combines results on-chip, avoiding extra result transfers. We apply this principle to linear layers and decode attention, while selective compensation limits added work when processing input prompts. On NVIDIA B200, compensated linear kernels using the NVFP4 format improve numerical precision while retaining latency close to uncompensated execution for the evaluated operations with few rows. In the evaluated NVFP4 inference configurations, ResServe adds at most 3.4% request-latency overhead over W4A4 and reduces request latency by 22.4-53.3% relative to W4A8. Across five models, code-generation accuracy improves over same-format uncompensated inference, with mean relative gains of 15.04% for NVFP4 and 53.57% for MXFP4.

cs.DC

新论文 · v1arXiv:2611.08660

AutoParallel: Composable, PyTorch-Native Automatic Parallelism for Production Training

Kaijian Wang, Sanket Purandare, Aditya Venkataraman, Will Constable, Ivan Kobzarev, Elias Ellison, Simon Fan, Edward Z. Yang 等 12 位作者

原文摘要

Training large models requires multiple interacting forms of parallelism. As models and clusters scale, finding and maintaining efficient parallelization plans becomes essential for training performance. Diverse model architectures, workloads, and hardware topologies demand flexible strategies, yet production PyTorch systems typically rely on hand-written recipes that are costly to adapt and tend to reuse familiar patterns, which may be suboptimal for a given workload. We present AutoParallel, a PyTorch-native planner for distributed model training. AutoParallel searches parallelization choices jointly across the training computation, allowing established techniques to be combined and unconventional strategies to emerge. It expresses its decisions using native PyTorch abstractions and produces ordinary PyTorch programs, avoiding a separate execution runtime while composing directly with the PyTorch compilation and training stack. AutoParallel also composes automatic decisions with model-specific distributed code, allowing dynamic or specialized regions to remain under user control. Plans are inspectable and replayable artifacts. Across dense, hybrid, and mixture-of-experts models on meshes of up to four dimensions and 128 NVIDIA H100 GPUs, AutoParallel matches or exceeds expert-written TorchTitan recipes on most configurations, with throughput gains of up to 18.3%, and discovers plans that uniform expert recipes do not express.

cs.DC

版本更新 · v2arXiv:2609.38377

PhyProbe: Rethinking Physical Consistency Evaluation in Generated Videos

Max Ku, Jiaojiao Fan, Zekun Hao, Francesco Ferroni, Heng Wang, Wenhu Chen, Ming-Yu Liu, Prithvijit Chattopadhyay

原文摘要

Evaluating the physical consistency of generated videos remains a fundamental challenge. Existing approaches rely on off-the-shelf vision-language models, which can often be myopic to physical dynamics, or fine-tuned evaluators trained on human annotations, which overfit to dataset-specific cues and fail to generalize. A key challenge is that existing supervision sources provide either relative ordering or absolute scores, but not both reliably and consistently across varied settings. To this end, we introduce PhyProbe, an evaluator that extracts features from a frozen pretrained spatio-temporal encoder and maps them to a scalar physical consistency violation score via a lightweight scoring head. PhyProbe is trained through a unified objective combining pairwise ranking, regression on noisy scalar annotations, and anchor-based calibration over a curated set of heterogeneous supervision sources. Experiments show that PhyProbe outperforms prior methods on most pairwise benchmarks spanning real-generated and generated-generated pairs under varying correspondence, with the largest gains in no-correspondence and generated-generated settings where existing fine-tuned evaluators degrade sharply. PhyProbe achieves strong correlation with human judgments, with close agreement between rank-based and linear metrics, indicating that scores are both well ordered and anchored to a stable [0, 1] scale. Further, despite being trained on supervision indicative of physical consistency, without explicit general-preference labels, PhyProbe also performs competitively on human preference benchmarks: consistent with the observation that physics violations are entangled with broader quality degradations.

cs.CV

新论文 · v1arXiv:2611.08472

XStack:机器学习仓库迁移基准

XStack: A Benchmark for ML Repo Migration

Linrong Cai, Taiming Lu, Haozhe Chen, Zhuang Liu

AI 速读 基于摘要

作者提出 XStack 基准,用于评估智能体在 PyTorch/NVIDIA GPU 与 JAX/Google TPU 两大生态之间迁移完整机器学习仓库的能力。

cs.LG · cs.AI

查看原文摘要

ML code should run where researchers choose, not where it was first written. Today, machine learning code mainly lives in two ecosystems, PyTorch on NVIDIA GPUs and JAX on Google TPUs. Migrating repositories between them means building and testing across frameworks and hardware. This migration is (1) a practical need, since it unlocks different hardware for researchers, and (2) a natural benchmark: it involves different frameworks and hardware, covers whole repositories, and is long-horizon and challenging. We introduce XStack, a benchmark that evaluates agents on migrating ML repositories between these two ecosystems. Each agent receives a PyTorch repository and a clean TPU machine and must deliver a JAX repository. Its output passes if it meets our grading criteria, based on the source repository's published result, authenticated and calibrated by us. Migration by agents remains unreliable: Claude Opus 5 passes 75%, GPT-5.6-sol 45%, and every other agent at most 40%. XStack is a step toward freeing ML code from its ecosystem. We release XStack to make progress toward reliable migration across frameworks and hardware measurable.