arXiv subjects

Dong Li

Publications and source records attributed to Dong Li.

At least 73 records · Page 4Linked to original sources

arXiv2026

Physics-Informed Neural Networks for Radial Consolidation of Combined Electroosmotic, Vacuum and Surcharge Preloading Considering Smear Effects

This study develops a dimensionless multi-domain physics-informed neural network (PINN) framework for electro-osmotic radial consolidation considering smear effects and combined vacuum and surcharge loading. Three PINN-based models are investigated: a standard soft-constrained PINN (Std-PINN), a modified gated PINN (Mod-PINN), and a modified gated PINN with hard-constraint boundary encoding (Mod-HC-PINN). The models are evaluated against FEM reference solutions under four loading cases, including constant vacuum, exponential vacuum, exponential vacuum with ramp surcharge, and exponential vacuum with cyclic haversine surcharge. The results indicate that the gated architecture applied in Mod-PINN improves the resolution of steep pressure gradients near the cathode and smear-zone interface under constant vacuum loading. Under time-dependent loading, the soft-constrained Mod-PINN shows reduced accuracy because it must learn multiple competing objectives simultaneously. The Mod-HC-PINN mitigates this issue by embedding the cathode boundary and initial conditions into the output structure, thereby reducing the optimization burden and improving physical consistency. The Mod-HC-PINN achieves MAE values of 0.43, 0.41, and 0.27 kPa for the exponential vacuum, ramp surcharge, and cyclic surcharge cases, respectively. Sensitivity analyses further demonstrate that the proposed framework remains robust across practical ranges of network architecture, collocation density, and permeability contrast.

cs.CE↗

arXiv2026

AgentKernelArena: Generalization-Aware Benchmarking of GPU Kernel Optimization Agents

GPU kernel optimization is increasingly critical for efficient deep learning systems, but writing high-performance kernels still requires substantial low-level expertise. Recent AI coding agents can iteratively read code, invoke compilers and profilers, and refine implementations, yet existing kernel benchmarks evaluate single LLM calls rather than full agent workflows, and none include both kernel-to-kernel optimization and unseen-configuration generalization testing. We present AgentKernelArena, an open-source benchmark for measuring AI coding agents on GPU kernel optimization. The benchmark contains 196 tasks spanning HIP-to-HIP optimization, Triton-to-Triton optimization, and PyTorch-to-HIP translation, and evaluates complete agent workflows in isolated workspaces using gated compilation, correctness, and performance checks, centralized scoring and an unseen-configuration generalization protocol that tests whether optimizations transfer to input configurations the agent never observed. Across production agents including Cursor Agent, Claude Code, and Codex Agent, we find near-perfect compilation and high correctness rates on most task categories, with the strongest configurations achieving mean speedups of up to 6.89x on PyTorch-to-HIP, 6.69x on HIP-to-HIP, and 2.13x on Triton-to-Triton tasks. Our unseen-configuration evaluation shows that HIP-to-HIP and Triton-to-Triton optimizations largely transfer to unseen input shapes, while PyTorch-to-HIP exhibits substantial correctness drops, indicating that agents generating kernels from scratch frequently hardcode shape-specific assumptions. AgentKernelArena is designed as a modular, extensible framework for rigorous evaluation of agentic GPU kernel optimization across agents, tasks, and hardware targets.

cs.CL↗

arXiv2026

HiDream-O1-Image: A Natively Unified Image Generative Foundation Model with Pixel-level Unified Transformer

The evolution of visual generative models has long been constrained by fragmented architectures relying on disjoint text encoders and external VAEs. In this report, we present HiDream-O1-Image, a natively unified generative foundation model via pixel-space Diffusion Transformer, that pioneers a paradigm shift from modular architectures to an end-to-end in-context visual generation engine. By mapping raw image pixels, text tokens, and task-specific conditions into a single shared token space, HiDream-O1-Image achieves a structural unification of multimodal inputs within an Unified Transformer (UiT) architecture. This native encoding paradigm eliminates the need for separate VAEs or disjoint pre-trained text encoders, allowing the model to treat diverse generation and editing tasks as a consistent in-context reasoning process. Extensive experiments show that HiDream-O1-Image excels across various generation tasks, including text-to-image generation, instruction-based editing, and subject-driven personalization. Notably, with only 8B parameters, HiDream-O1-Image (8B) achieves performance parity with or even surpasses established state-of-the-art models with significantly larger parameters (e.g., 27B Qwen-Image). Crucially, to validate the immense scalability of this paradigm, we successfully scale the architecture up to over 200B parameters. Experimental results demonstrate that this massive-scale version HiDream-O1-Image-Pro (200B+) unlocks unprecedented generative capabilities and superior performance, establishing new state-of-the-art benchmarks. Ultimately, HiDream-O1-Image highlights the immense potential of natively unified architectures and charts a highly scalable path toward next-generation multimodal AI.

cs.CV↗

arXiv2026

ChunkFlow: Communication-Aware Chunked Prefetching for Layerwise Offloading in Distributed Diffusion Transformer Inference

Layerwise offloading reduces the GPU memory footprint of large diffusion transformer (DiT) inference by prefetching upcoming layers from host memory, but its effectiveness hinges on hiding prefetch latency behind per-layer computation. This assumption breaks down when the per-GPU compute workload is small. Moreover, on PCIe-only nodes, prefetch and inter-GPU collective communications such as all-reduce and all-to-all contend on the shared PCIe path, exposing prefetch latency even when compute would otherwise hide it. We revisit layerwise offloading as a co-scheduling problem between prefetch and communication, guided by a first-order analytical model that predicts when prefetch can be hidden by computation. Building on this model, we design ChunkFlow, a communication-aware, chunk-granular offloading runtime that adaptively yields to collective communication and smoothly trades GPU memory for prefetch volume. On three representative diffusion transformers running on two H100 GPUs over PCIe with Ulysses sequence parallelism, ChunkFlow delivers up to 1.28x step-time speedup over SGLang's existing layerwise offloading, reduces peak GPU memory by up to 49% over the no-offload baseline at near-identical step time once the workload is large enough, and exposes a tunable memory-latency tradeoff that recovers near-zero step-time overhead in the small-workload regime.

cs.DC↗

arXiv2026

Theory-optimal Quantization Based on Flatness

Post-training quantization has emerged as a widely adopted technique for compressing and accelerating the inference of Large Language Models (LLMs). The primary challenges in LLMs quantization stem from activation outliers, which significantly degrade model performance especially at lower bit precision. While recent approaches attempt to mitigate outliers through linear transformations across feature dimensions, our analysis reveals that the transformed weights and activations still exhibit persistent outlier patterns with concentrated magnitude distributions. In this paper, we first model the mathematical relationship between quantization error and outliers, and then introduce a new metric Flatness to quantify the distribution of outliers. Based on this, we derive the theoretical optimal solution with respect to Flatness. Building on these insights, we propose Bidirectional Diagonal Quantization (BDQ), a novel post-training quantization framework that effectively disperses outlier patterns through optimized matrix transformations. BDQ strategically distributes outlier magnitudes across matrix dimensions via learned diagonal operations. Extensive experiments demonstrate that BDQ establishes a new quantization benchmark. It achieves less than 1\% accuracy drop in W4A4 quantization on the LLaMA-3-8B model. In the more challenging W2A4KV16 experiment, compared to state-of-the-art approaches, BDQ reduces the performance gap by 39.1\% on the DeepSeek-R1-Distill-LLaMA-70B model.

cs.LG↗

arXiv2026

PARD-2: Target-Aligned Parallel Draft Model for Dual-Mode Speculative Decoding

Speculative decoding accelerates Large Language Models (LLMs) inference by using a lightweight draft model to propose candidate tokens that are verified in parallel by the target model. However, existing draft model training objectives are not directly aligned with the inference-time goal of maximizing consecutive token acceptance. To address this issue, we reformulate the draft model optimization objective, shifting the focus from token prediction accuracy to the overall acceptance length. In this paper, we build upon PARD to propose PARD-2, a dual-mode speculative decoding framework with Confidence-Adaptive Token (CAT) optimization. This approach adaptively reweights each token to better align with the verification process. Notably, PARD-2 enables a single draft model to support both target-dependent and target-independent modes. Experiments across diverse models and tasks demonstrate that PARD-2 achieves up to 6.94$\times$ lossless acceleration, surpassing EAGLE-3 by 1.9$\times$ and PARD by 1.3$\times$ on Llama3.1-8B. Our code is available at https://github.com/AMD-AGI/PARD.

cs.CL↗

arXiv2026

CCCL: Node-Spanning GPU Collectives with CXL Memory Pooling

Large language models (LLMs) training or inference across multiple nodes introduces significant pressure on GPU memory and interconnect bandwidth. The Compute Express Link (CXL) shared memory pool offers a scalable solution by enabling memory sharing across nodes, reducing over-provisioning and improving resource utilization. We propose \name, a collective communication library, leveraging the CXL shared memory pool to support cross-node GPU operations without relying on traditional RDMA-based networking. Our design addresses the challenges on synchronization, data interleaving, and communication parallelization faced by using the CXL shared memory pool for collective communications. Evaluating on multiple nodes with a TITAN-II CXL switch and six Micron CZ120 memory cards, we show that \name achieves highly efficient collective operations across hosts, demonstrating CXL's potential for scalable, memory-centric GPU communication. Our evaluation demonstrates that \name achieves average performance improvements of 1.34$\times$ for AllGather, 1.84$\times$ for Broadcast, 1.94$\times$ for Gather, and 1.04$\times$ for Scatter, compared to the original RDMA-based implementation over 200 Gbps InfiniBand. \textcolor{dong}{In addition, the evaluation with a case of LLM training shows 1.11$\times$ speedup compared with the InfiniBand while saving production cost by $2.75\times$ in hardware.}

cs.DC↗

arXiv2026

FluxFlow: Conservative Flow-Matching for Astronomical Image Super-Resolution

Ground-to-space astronomical super-resolution requires recovering space-quality images from ground-based observations that are simultaneously limited by pixel sampling resolution and atmospheric seeing, which imposes a stochastic, spatially varying PSF that cannot be resolved through upsampling alone. Existing methods rely on synthetic training pairs that fail to capture real atmospheric statistics and are prone to either over-smoothed reconstructions or hallucination sources with no physical counterpart in the observed sky. We propose FluxFlow, a conservative pixel-space flow-matching framework that incorporates observation uncertainty and source-region importance weights during training, and a training-free Wiener-regularized test-time correction to suppress hallucination sources while preserving recovered detail. We further construct the DESI--HST Dataset, the large-scale real-world benchmark comprising 19,500 real co-registered ground-to-space image pairs with real atmospheric PSF variation. Experiments demonstrate that FluxFlow consistently outperforms existing baseline methods in both photometric and scientific accuracy.

cs.CV↗

arXiv2026

Differential Spatial Modulation with Transmit Diversity for Pinching-Antenna Systems

Pinching antenna (PA) systems provide a new spatial degree of freedom by flexible activation of pinching positions. However, the resulting effective channel strongly depends on the activated pinching positions, rendering conventional coherent transmission generally relies on accurate acquisition of instantaneous channel state information (CSI) and incurring substantial pilot overhead. To address this challenge, we propose a differential spatial modulation (DSM) scheme for PA systems, termed as DSM-PA. Specifically, a differential transmission scheme with an embedded Alamouti coding structure is designed, where information bits are conveyed via phase variations between adjacent symbol blocks. This design enables noncoherent transmission without requiring instantaneous CSI while simultaneously achieving transmit diversity. Moreover, to fully exploit the spatial degrees of freedom of PA systems, a pinching position-based index modulation (IM) rule is developed to enhance spectral efficiency. An asymptotically tight upper bound on the average bit error rate (BER) over quasi-static Rician fading channels is derived using the moment-generating function (MGF) method. The diversity analysis also reveals that the proposed DSM-PA scheme achieves full transmit diversity. Finally, simulation results verify the accuracy of the BER analysis and demonstrate the effectiveness of the proposed DSM-PA scheme.

eess.SP↗

arXiv2026

Distributed Generative Inference of LLM at Internet Scales with Multi-Dimensional Communication Optimization

Decentralized LLM inference distributes computation among heterogeneous nodes across the internet, offering a performant and cost-efficient solution, alternative to traditional centralized inference. However, the low cross-node network bandwidth makes communication the primary bottleneck. In this paper, we introduce BloomBee, an internet-scale distributed LLM inference framework. BloomBee integrates LLM-layer assignment, micro-batching and tensor offloading to optimize communication from multiple dimensions. Additionally, BloomBee formulates the coordination of these techniques as an optimization problem and solves it using dynamic programming. BloomBee also customizes lossless compression and speculative decoding according to low-bandwidth network settings to reduce communication overhead. We evaluate BloomBee across a spectrum of network environments and show that it improves service throughput by up to 1.76x. It also reduces average latency by up to 43.20% compared to state-of-the-art decentralized LLM inference systems. BloomBee is open-sourced.

cs.DC↗

arXiv2026

Training-Free Loosely Speculative Decoding: Accepting Semantically Correct Drafts Beyond Exact Match

Large language models (LLMs) achieve strong performance across diverse tasks but suffer from high inference latency due to their autoregressive generation. Speculative Decoding (SPD) mitigates this issue by verifying candidate tokens in parallel from a smaller draft model, yet its strict exact-match verification discards many semantically valid continuations. Moreover, existing training-based SPD methods often suffer from performance degradation on out-of-distribution (OOD) tasks. To this end, we propose Training-Free Loosely Speculative Decoding (FLy), a novel method that loosens the rigid verification criterion by leveraging the target model's self-corrective behavior to judge whether a draft-target mismatch remains semantically valid. FLy introduces a two-tier mechanism: an entropy-level gate that identifies whether the current token allows multiple plausible alternatives or is nearly deterministic, and a token-level deferred window that distinguishes genuine errors from differently worded yet semantically correct variants. To further reduce latency, we design a multi-level acceleration strategy that accelerates not only the target model but also the drafter itself. Owing to its training-free design, FLy composes seamlessly with arbitrary draft-target pairs and generalizes across models and domains without hyperparameter re-tuning. Experiments show that FLy preserves more than 99% of the target model's accuracy while achieving an average 2.81x speedup on Llama-3.1-70B-Instruct and 5.07x speedup on the 405B variant. Notably, on out-of-domain datasets, our method remains highly effective and outperforms the training-based method EAGLE-3 by 1.62x.

cs.CL↗

arXiv2026

SVOM/VT: Instrument Overview, Science Objectives, and First-Year Performance

The 44-cm Visible Telescope (VT) aboard the Space-based Variable Objects Monitor (SVOM) is a dual-band (400-650 nm and 650-1000 nm) instrument designed to detect and characterize the optical counterparts of gamma-ray bursts (GRBs) and other high-energy transients. This paper presents the VT's design, scientific objectives, observing strategies, and both space- and ground-based data processing pipelines, along with its first-year in-orbit performance. In-orbit commissioning tests confirm a sensitivity of 22.5 AB mag (300 s exposure), extendable to $\sim\!24$ AB mag through stacking. This performance enables the VT to monitor over 100 GRBs in its first year with an exceptional $\sim\!80\%$ detection rate for \textit{SVOM}/ECLAIRS-triggered bursts and ToO-observed bursts from other missions (e.g., \textit{Swift, Fermi, Einstein Probe (EP)}), outperforming \textit{Swift}/UVOT's $\sim\!40\%$ detection rate. Beyond its exceptional detection efficiency, the VT played a key role in identifying high-redshift GRBs-most notably GRB 250314A (z = 7.3). Its deep upper limits at long wavelengths (up to 1 $μ$m) were pivotal in guiding follow-up observations with large ground-based telescopes, enabling crucial near-infrared (NIR) detections. With its rapid response, deep sensitivity, and real-time processing capabilities, the VT is a key instrument for GRB research in \textit{SVOM}-era, enabling critical studies of GRB optical afterglows, circumburst environments, relativistic jet dynamics, and the origins of optically dark bursts.

astro-ph.HE↗

arXiv2026

SVOM/VT: Real-Time Onboard Data Processing

The SVOM Visible Telescope (VT) is critical for the rapid identification of gamma-ray burst (GRB) optical counterparts, particularly for high-redshift candidates that require immediate infrared spectroscopic follow-up. To address the stringent bandwidth constraints of the VHF downlink while ensuring real-time data availability, we developed the VT Onboard Data Processing Pipeline (VOPP).This paper details the software architecture, algorithms, and hardware implementation of VOPP using an FPGA and a CPU. The pipeline performs essential real-time tasks, including image quality assessment, dark and flat-field correction, and optimized image stacking to mitigate cosmic ray contamination and variable background noise. Furthermore, it generates compact source catalogs and highly compressed 1-bit images to facilitate rapid downlink.In-flight performance analysis confirms the pipeline's robustness, demonstrating the availability of VT VHF data for 78 percent of promptly slewed SVOM GRBs, with 56 percent leading to the identification of optical counterparts, typically within 18 minutes post-trigger.

astro-ph.IM↗

arXiv2026

NeuronMLP: Efficient LLM Inference via Singular Value Decomposition Compression and Tiling on AWS Trainium

Emerging AI accelerators have started to gain attention and offer new opportunities for efficient inference of large language models (LLMs). Trainium, an AI accelerator recently developed by Amazon Web Services (AWS), provides an attractive option for LLM inference through its heterogeneous architecture. However, leveraging Trainium architecture for high performance can be challenging because of its systolic array architecture and special requirement on data layout. In this paper, we propose NeuronMLP, an efficient LLM inference method based on Singular Value Decomposition (SVD) compression and tiling on AWS Trainium. We introduce a series of techniques customized to Trainium based on kernel fusion and novel caching strategies to reduce data movement across the software-managed memory hierarchy, maximize SRAM bandwidth, and avoid expensive matrix transpose. The proposed method is specifically optimized for multi-layer perceptron (MLP) layers in LLMs, which serve as a critical computational kernel for inference on Trainium. Evaluating on nine datasets and six recent LLMs, we show that NeuronMLP significantly outperforms the state-of-the-art Neuron Kernel Interface (NKI)-based matrix multiplication (matmul) kernel implemented by AWS on Trainium: at the kernel level, it achieves an average 1.35x speedup, which translates to an average 1.21x speedup for end-to-end LLM inference, under a compression ratio of 0.05.

cs.CL↗

arXiv2026

CoCoDiff: Optimizing Collective Communications for Distributed Diffusion Transformer Inference Under Ulysses Sequence Parallelism

Diffusion Transformers (DiTs) are increasingly adopted in scientific computing, yet growing model sizes and resolutions make distributed multi-GPU inference essential. Ulysses sequence parallelism scales DiT inference but introduces frequent all-to-all collectives that dominate latency. Overlapping these with computation is difficult due to tight data dependencies, large message volumes, and asymmetric interconnect bandwidths. We introduce CoCoDiff, a distributed DiT inference engine exploiting two observations: (1) V requires only linear projection while Q/K need additional normalization and RoPE, creating opportunities to overlap V's communication with Q/K computation; (2) adjacent denoising steps produce similar tensors, yielding temporal redundancy. CoCoDiff introduces three mechanisms: Tile-Aware Parallel All-to-all (TAPA) decomposes collectives into topology-aligned phases; V-First scheduling hides V's communication behind Q/K computation; and V-Major selective communication transmits only active projections on slow interconnects. On the Aurora supercomputer with four DiT models across 1-8 nodes (up to 96 Intel GPU tiles), CoCoDiff achieves an average speedup of 3.6x, peaking at 8.4x.

cs.DC↗

arXiv2026

IR-Flow: Bridging Discriminative and Generative Image Restoration via Rectified Flow

In image restoration, single-step discriminative mappings often lack fine details via expectation learning, whereas generative paradigms suffer from inefficient multi-step sampling and noise-residual coupling. To address this dilemma, we propose IR-Flow, a novel image restoration method based on Rectified Flow that serves as a unified framework bridging the gap between discriminative and generative paradigms. Specifically, we first construct multilevel data distribution flows, which expand the ability of models to learn from and adapt to various levels of degradation. Subsequently, cumulative velocity fields are proposed to learn transport trajectories across varying degradation levels, guiding intermediate states toward the clean target, while a multi-step consistency constraint is presented to enforce trajectory coherence and boost few-step restoration performance. We show that directly establishing a linear transport flow between degraded and clean image domains not only enables fast inference but also improves adaptability to out-of-distribution degradations. Extensive evaluations on deraining, denoising and raindrop removal tasks demonstrate that IR-Flow achieves competitive quantitative results with only a few sampling steps, offering an efficient and flexible framework that maintains an excellent distortion-perception balance. Our code is available at https://github.com/fanzh03/IR-Flow.

cs.CV↗

arXiv2026

Spatio-temporal Characteristics of Very Long-periodic Pulsations in Solar Metrewave Bursts: Implications for their Origins

We traced the origin of very long-periodic pulsations (VLPs) in type-I burst chains on 2024 February 14. Seven successive and repetitive pulsation structures appeared in radio dynamic spectra in the metric waveband, which were simultaneously measured by CBSm, DART, and MUSER-L. A quasi-period at about 160$^{+11}_{-6}$ s, determined by the fast Fourier transform, was detected in the frequency range of about 210-280 MHz. Imaging observations from DART and SDO reveal that the type-I burst chains occur above two groups of sunspot umbrae connected by coronal loops. A quasi-period of approximately 170 s was also identified in the sunspot umbrae and coronal loops. The burst chains exhibit strong circular polarization and high brightness temperature, and they show spatiotemporal correlation with emerging magnetic flux. The number densities at the loop top and double footpoints can produce radio emission and generate type-I burst chains in the frequency range of 210-280 MHz. Our observations support the scenario that plasma emission serves as the primary generation mechanism of type-I bursts, with VLPs most likely being modulated by the slow magnetoacoustic waves originating from sunspot umbrae. The observed frequency drift of burst chains may reflect the density attenuation along coronal loops.

astro-ph.SR↗

arXiv2026

Ego-InBetween: Generating Object State Transitions in Ego-Centric Videos

Understanding physical transformation processes is crucial for both human cognition and artificial intelligence systems, particularly from an egocentric perspective, which serves as a key bridge between humans and machines in action modeling. We define this modeling process as Egocentric Instructed Visual State Transition (EIVST), which involves generating intermediate frames that depict object transformations between initial and target states under a brief action instruction. EIVST poses two challenges for current generative models: (1) understanding the visual scenes of the initial and target states and reasoning about transformation steps from an egocentric view, and (2) generating a consistent intermediate transition that follows the given instruction while preserving object appearance across the two visual states. To address these challenges, we propose the EgoIn framework. It first infers the multi-step transition process between two given states using TransitionVLM, fine-tuned on our curated dataset to better adapt to this task and reduce hallucinated information. It then generates a sequence of frames based on transition conditions produced by the proposed Transition Conditioning module. Additionally, we introduce Object-aware Auxiliary Supervision to preserve consistent object appearance throughout the transition. Extensive experiments on human-object and robot-object interaction datasets demonstrate EgoIn's superior performance in generating semantically meaningful and visually coherent transformation sequences.

cs.CV↗