Distributed Inference

Follow this topic Ask about this topic

Momentum

8 papers in the last four weeks, against 2 the four weeks before. 0.1% of all new papers.

Jul 13Week of Sep 28

Latest papers 55

Sep 30, 2026cs.DC

Cascadia: A Control-Plane-Free Alternative to Hyperconverged AI Infrastructure

We present Cascadia, a system for serving large language models on fleets of commodity Intel AIPCs using their CPU, integrated-GPU, and NPU resources. Every node embeds ingress, scheduling, and execution; inference requests require no dedicated routing control plane. Nodes join a libp2p QUIC mesh using CA-issued ed25519 admission certificates, gossip signed capabilities, exchange live load over direct peer streams, and route OpenAI-compatible requests to eligible peers. An operator-run certificate authority handles admission and fleet management outside the inference path. Three serving modes share one interface: whole-model execution on one node, load-balanced replicas, and pipeline-sharded chains using the compilation and speculative decoding mechanism of our companion paper. Optional KV-cache mobility reuses compatible conversation prefixes after a routing move, with cold recomputation on a miss. Signed response receipts and hash-chained logs support provenance and audit. A three-node Phi-3.5-mini NPU testbed delivered 3.10x the response throughput of its one-node configuration under ten concurrent requests; a separate four-node deployment recorded 4.06x the throughput of direct single-node serving. Paired latency observations, runtime measurements, and internal functional checks characterize the tested configurations. We compare Cascadia with IBM, Nutanix, VMware, and HPE platforms on deployment footprint, hardware requirements, scheduling, scaling, licensing, and trust, using vendor documentation. The paper repository provides benchmark scripts, curated measurements, and a claim-to-evidence map.

Matias Parij, Pawan Paudel, Tate Berenbaum +1

Not Community Labs Inc. · Intel Corporation

LLM Serving LLM Inference Scheduling Distributed Inference

Sep 29, 2026cs.DC

Purlin: Separating Orchestration from the Datapath of Collectives

Distributed inference depends on GPU collective communication that must keep pace with evolving hardware and specialized workloads. However, existing collective implementations often couple semantics, orchestration (where and when data moves), and the datapath (how data moves). This coupling makes it costly to adopt new hardware mechanisms and customize communication for applications. We present Purlin, a scale-up communication framework that separates these concerns. At the top of Purlin, we specify collectives as a naming of an input and output layout and a copy or reduction operation. In the middle, we introduce a shared orchestration protocol, Stage, Notify, And Consume (SNAC), which derives coordination from these specifications. Below SNAC sits a hardware-specific datapath we call Atom, which implements two key data movement primitives for collectives: copy and reduce. This separation lets us customize collectives and adopt new hardware mechanisms while reusing orchestration via SNAC. We evaluate Purlin on A100, H200, and B200 GPUs. Across seven collectives, Purlin achieves latency speedups of up to 5.14x and bandwidth improvements of up to 4.50x over baselines. Integrated into SGLang, Purlin improves offline LLM serving throughput and interactivity by 1.13x on average and up to 1.37x over baselines. For online LLM inference, Purlin improves interactivity by 1.26x on average and up to 2.85x, with the largest gain occurring under overload. For diffusion image generation, Purlin reduces end-to-end latency by up to 1.13x.

Osayamen Jonathan Aimuyo, Swapnil Gandhi, Christos Kozyrakis

Stanford University · NVIDIA & Stanford University

LLM Inference Efficiency High-Performance Computing Distributed Inference

Sep 29, 2026eess.IV

Reliability Testing of Medical Model Performance under Distributed Deployment

Distributed inference has become an indispensable part of deploying medical models under practical latency, memory, and throughput constraints. Although modern frameworks improve serving efficiency through tensor parallelism, mixed precision, kernel fusion, and multi-device communication, they are generally assumed to preserve the behavior observed during centralized HuggingFace evaluation. This assumption creates an evaluation-deployment mismatch: a model may pass offline evaluation but produce a different output after the execution stack changes. To address this mismatch, we propose a testing framework and an improved, distributed-execution-sensitive medical-model benchmark that evaluates the same checkpoint and input under a centralized HuggingFace reference and matched distributed deployments. Extensive experiments across language, vision, and multimodal medical models show that execution changes can produce measurable output disagreements. Across supported visual settings, the test success rate ranges from 0.21 to 0.43 for single-modality models and from 0.32 to 0.98 for multimodal models. The benchmark is aimed at extending medical-model evaluation from capability and security to evaluation-deployment consistency.

Yifei Wang, Xiaohan Zhang, Youtao Ding +4

Shanghai Jiao Tong University, China · Beihang University, China · Nanyang Technological University, Singapore +1

Multimodal Model Evaluation Distributed Inference

Sep 28, 2026cs.DC

Kafila: Serving Large Language Models on a Trusted Set of Heterogeneous Commodity Machines

Between them, the members of a research group or a circle of friends own several consumer computers, none large enough to run a capable large language model. Existing systems pool such capacity across open swarms anyone may join, which a group admitting only trusted machines cannot use. Bounding membership removes what they depend on: a swarm holds each part of the model on several peers and routes around a slow one. A bounded session must use every device it admits. Its pipeline advances at the pace of whichever device received a share it cannot serve quickly, so the division has to be right before serving begins. We propose Kafila, whose protocol assembles a ring from behind NATs, preferring direct paths and relaying where traversal fails, while its planner measures each device's memory bandwidth, capacity and reachability, divides the model exactly for a fixed ring order, and places the head, which holds the embedding and output projection, together with that division rather than beforehand. On machines with different capabilities across three fleets, from a shared LAN to five devices spanning two continents, Kafila shortens the slowest pipeline stage by up to

5.2\times

against the even split of pipeline parallelism, as in GPipe, and up to

3\times

against the memory-proportional split of personal-device inference, as in exo, keeps 75 to 87 per cent of the committed hardware doing work where those divisions fall below half, and serves a model no uniform split can place on the fleet at all. What that is worth to a user depends on how much of a token is computation rather than network. Where the members share a network the same division returns

1.56\times

the throughput of a uniform split and

1.25\times

of a memory-proportional one, and under four concurrent users that lead compounds to

3.2\times

rather than fading, each user served at almost the rate of one.

Murtaza Rangwala, Richard O. Sinnott, Rajkumar Buyya

Quantum Cloud Computing and Distributed Systems (qCLOUDS) Lab, School of Computing and Information Systems, The University of Melbourne, Australia

LLM Inference LLM Inference Scheduling Heterogeneous Computing Distributed Inference

Sep 16, 2026cs.CV

vidax: A Unified JAX Framework for Video Generative Models on Accelerator Meshes

Open-source video generative models ship almost exclusively as PyTorch/CUDA reference implementations. This leaves Cloud TPU pods without a production-ready inference path, despite offering large, cost-effective accelerator memory pools ideal for long-sequence spatiotemporal attention. We present vidax, an open-source JAX/Flax inference engine and zero-copy PyTorch-to-JAX weight translator for modern video generation architectures. vidax covers a diverse set of spatiotemporal models --- including Diffusion Transformers, omnimodal Mixture-of-Transformers, 3D VAEs, text encoders, and native samplers --- with zero PyTorch dependency in the execution path. The framework unifies 1D tensor parallelism with DeepSpeed-Ulysses sequence parallelism on a single JAX sharding mesh, integrates TPU flash-attention kernels, and implements per-layer weight offloading to support reference resolutions that exceed single-device memory. We benchmark compile times, latency, and peak memory utilization on TPU v4-8 hardware, and document real-world numerical bugs surfaced during checkpoint translation. vidax is released open-source as a baseline for JAX and TPU video generation research.

Congyue Deng

Massachusetts Institute of Technology

Video Diffusion Models Tensor Parallelism Video Generation Distributed Inference Diffusion Model Inference Acceleration

Sep 15, 2026cs.LG

Recovering Governing Dynamics from Distributed Observations via Exact Spline Merging

Scientific observations are frequently distributed across locations, time periods, and institutions. Combining such observations into a continuous, differentiable field enables recovering governing physical parameters from its derivatives. This paper makes two contributions in this setting. First, the established additive structure of fixed-basis ridge-regression statistics is applied to tensor-product spline fields: each data holder computes a local Gram matrix and moment vector, and the merged solution is mathematically identical to centralized fitting, with no raw data shared and no iterative synchronization. This property is specific to the fixed-feature squared-error setting; the present derivation does not establish an analogous guarantee for general jointly trained multilayer networks. Second, a complete pipeline connects distributed observations to physical parameter inference through field reconstruction, derivative extraction, and linear regression. The pipeline is validated on four PDEs: diffusion, wave, heat-with-source, and the nonlinear viscous Burgers equation, recovering governing parameters to sub-percent accuracy in the linear cases and 5% for Burgers. In all cases, distributed merging introduces zero degradation relative to centralized fitting. Application to 41 years of NOAA sea-surface temperature data confirms the result on real spatiotemporal observations. Source code to reproduce all experiments is available at https://github.com/NAVEENMN/gramfield.

Naveen Mysore

University of California, Santa Barbara · Dyssonance AI

PDE Discovery Distributed Inference

Sep 9, 2026cs.DC

Composable CXL Memory as a Kubernetes-Native Shared Memory for LLM Serving

We present a Kubernetes Dynamic Resource Allocation (DRA) driver that makes composable CXL memory a schedulable cluster resource, and evaluate the resulting shared-memory tier for cross-node KV-cache reuse in LLM serving. The driver composes CXL regions on demand, materializes them as DAX devices on each participating host, and injects them into pods under a single Container Device Interface (CDI) name so that pods on different nodes access the same physical region. A shared-memory connector for vLLM/llm-d uses that region as a KV-cache tier with a slot directory embedded inside the shared medium, which eliminates the need for an external metadata service. On a two-node cluster with a 512,GiB CXL appliance and Qwen2.5-7B-Instruct, cross-node prefix reuse reduces TTFT by 5.5

\times

--36.6

\times

at an external hit rate of 95.4--99.5,%, while node-local tiers (GPU prefix caching, CPU-DRAM offload) fall back to full recompute. The sharing gap, defined as the latency ratio between cross-node and same-node reuse, is 1--4%, indicating that cross-node reuse incurs little additional latency relative to same-node reuse on our testbed. Both replicas run full engines; the study demonstrates memory disaggregation rather than prefill/decode disaggregation. We report this as a feasibility study rather than a performance evaluation.

Hongjian Fan, Kevin Zhang, David Habinsky +1

Research Group, Seagate Technology, LLC

LLM Inference Efficiency KV Caching KV-Cache Reuse Distributed Inference

Sep 7, 2026cs.IT

A Fundamental Limit in Decentralized Decision-Making

In decentralized decision-making, several agents connected according to a network graph aim at solving a classification problem by collecting streaming observations. Due to decentralization, they run an iterative algorithm where, at each iteration, they can only exchange information locally with their neighbors. While decentralized estimation solutions have been shown to match the performance of optimal centralized systems, we show here that surprisingly this conclusion does not hold for decentralized decision-making. Specifically, we prove that the error probability for the best decentralized decision strategy exhibits an irreducible loss with respect to the optimal centralized classifier. This result establishes a fundamental limit for the performance of any decentralized decision strategy. We obtain an analytical relation showing that this limit is related to the interplay between decentralization and classification. The first aspect appears through the distances between the nodes in the graph, while the second aspect plays through the moment generating functions of the likelihood ratios that describe the decision problem. By applying the derived closed-form relation to different network topologies and inference problems, we observe some interesting and perhaps unexpected behavior emerging. In particular, we characterize the scaling law (with the network size) for the loss over popular network topologies, showing that the error probabilities might differ by orders of magnitude; and we examine how performance is affected by the relative distance between informative and uninformative agents over the graph.

Marco Carpentiero, Felice Scala, Vincenzo Matta +1

Department of Information and Electrical Engineering and Applied Mathematics (DIEM), University of Salerno, I-84084 Fisciano (SA), Italy · Institute of Electrical and Micro Engineering, EPFL, CH-1015 Lausanne, Switzerland

Classification Distributed Inference Information-Theoretic Lower Bounds

Sep 1, 2026cs.LG

Manifold-Aware General Coded Computing for Straggler-Resilient Distributed Computing

Existing coded-computing designs do not explicitly exploit the intrinsic structure of the input data. In communication systems, statistical structure and redundancy are often removed through source coding (or compression) before channel coding is applied. This principle, however, does not transfer directly to coded computation. In many computational tasks, particularly in machine learning, the structure of the data is precisely what the computation seeks to exploit to infer outputs or learn meaningful patterns. Consequently, coded-computing schemes should preserve and leverage this structure in their code design, rather than ignoring or eliminating it through source coding. This observation motivates a different perspective on code construction. In many channel-coding schemes, such as Reed-Solomon codes, coded symbols are generated by evaluating a low-dimensional algebraic representation at selected points. In contrast, many high-dimensional datasets naturally concentrate near low-dimensional manifolds. In this paper, we exploit this intrinsic geometry by designing coded samples that follow the natural manifold of the data, rather than imposing an artificial low-dimensional structure unrelated to the data distribution. Inspired by graph-based manifold learning, we propose a manifold-aware encoding strategy for general coded computing (GCC). Experiments on neural network inference and high-dimensional polynomial evaluation demonstrate that the proposed strategy consistently and significantly reduces the mean squared recovery error under straggling compared with standard GCC.

Parsa Moradi, Mohammad Ali Maddah-Ali

University of Minnesota, Twin Cities Minneapolis, MN, USA

Distributed Inference

Aug 12, 2026cs.DC

User-Assisted Collaborative Distributed Inference for Efficient QoS-Aware Autoscaling

Growing demand for artificial intelligence (AI) inference services requires scalable infrastructure, yet centralized serving costs rise with demand. We propose a collaborative distributed inference system combining dedicated infrastructure with resources contributed by service users. Dedicated resources provide baseline capacity for maintaining quality of service (QoS), while volunteered resources absorb increasing demand without proportional growth in centralized infrastructure. To capture stochastic and dynamic interactions among users, resources, tasks, and policies, we develop a high-dimensional generative Markov model with structured temporal factorization. The model supports simulation and provides a foundation for task scheduling and QoS-aware resource allocation optimization. We evaluate the system across user populations, resource capacities, and centralized and distributed scheduling policies. Simulations show that distributed scheduling becomes increasingly advantageous as the user population grows, improving request completion and P99 latency while substantially reducing dedicated resource consumption. These results demonstrate the feasibility of user-assisted collaborative inference for infrastructure-efficient autoscaling.

Alfreds Lapkovskis, Ali Beikmohammadi, Sindri Magnússon +1

Department of Computer Systems and Sciences, Stockholm University SE-106 91 Stockholm, Sweden

Distributed Inference Online Resource Allocation

Aug 1, 2026cs.NI

HetRoute Heterogeneous and Cost-aware Collaborative Routing Framework for Distributed Edge MoE Inference

Mixture-of-Experts (MoE) models have become a dominant architecture for large-scale AI services, yet deploying them over geo-distributed heterogeneous edge servers remains challenging. When the Top-k activated experts of a token are spread across multiple servers, the optimal routing depends jointly on cross-server link bandwidth, heterogeneous GPU computing capability, GPU-CPU expert loading delay, instantaneous queueing backlog, and replica-level quantization quality loss. Existing distributed inference and MoE serving methods address these factors separately and do not provide a unified framework for online multi-server collaborative routing. In this paper, we propose HetRoute, a heterogeneous-cost-aware collaborative routing framework for distributed edge MoE inference. HetRoute introduces a unified per-assignment cost model that explicitly captures four cost components: cross-server transmission, GPU-CPU offloading, GPU computation with queueing, and quantization-induced quality penalty. Guided by this model, the offline stage determines expert server placement, GPU-CPU residency, and replica precision through a routing-cost-coupled deployment algorithm, while the online stage routes the Top-k activated expert set as a whole by minimizing the bottleneck layer cost via exact enumeration or beam search. Theoretical analysis establishes fallback feasibility, a bound on the number of participating servers, per-layer optimality for small candidate domains, and online computational complexity. Trace-driven evaluation on three MoE models over a heterogeneous 10-server edge testbed shows that HetRoute reduces average inference latency by up to 59.0% and P99 latency by up to 58.0%, cuts cross-server traffic by up to 72.1%, and achieves 2.13x throughput improvement compared with representative baselines, while keeping quality degradation within the configured budget.

Xin Yuan, Ning Li, Wenchao Xu +2

School of Ocean Engineering, Harbin Institute of Technology, Heilongjiang, China · School of Artificial Intelligence, University of Science and Technology Beijing, China · Division of Integrative Systems and Design, Hong Kong University of Science and Technology, Hong Kong +1

Edge Inference Cost-Aware Inference Mixture-of-Experts Inference Expert Routing Distributed Inference

Jul 27, 2026cs.NI

Denial of Deadline: Network-Driven Accuracy Collapse in Distributed Inference Pipelines

Inference systems increasingly combine a fast path that returns predictions within the application's latency deadline together with a higher-accuracy slow path that runs higher-compute methods on stronger, remote hardware, so its results can be returned on time and combined with the fast path predictions. Across several application domains, we abstract this inference architecture as a fast path, a slow path, and a coordination layer with two functions: a router that invokes the slow path and a merger that decides whether to incorporate its returned predictions. In this work, we show that this new coordination layer exposes a new attack surface: shaped workload attacks, e.g., Yo-Yo bursts, can exploit contention at shared resources along the slow path to push benign users' slow-path predictions past their latency deadlines. The merger then discards those predictions, while the fast path continues to return timely outputs. We refer to the resulting loss of slow-path accuracy benefits as accuracy collapse. We demonstrate accuracy collapse in a two-tier edge-cloud multi-object tracking pipeline in autonomous driving. In simulation, approximately 4,000 burst-shaped requests increase benign p99 latency from 92ms to 2s, nearly eliminating the benefit of the slow path's cloud inference, reducing object tracking quality by 7.0 HOTA points on average. We further find that accuracy degradation can significantly vary (2.0-18.7 HOTA points), depending on the video intervals that are targeted in the attack, and that certain rare classes (e.g., stop signs) lose nearly half of their pre-attack prediction accuracy. These results show that workload attacks can degrade prediction quality without needing either access to model weights or victim data, and motivate research on attacks and defenses for routing, merging, scheduling, and resource isolation in these emerging inference pipeline architectures.

Jhonatan Tavori, Gur-Eyal Sela, Ion Stoica +1

Columbia University · UC Berkeley

Adversarial Attacks Edge Computing Distributed Inference

Jul 25, 2026cs.DC

Gleam: Adaptive Network-Efficient CUDA API Remoting for Cross-Device GPU Sharing over LANs

This paper aims to enable computation- and communication-efficient GPU sharing across devices within local area networks (LANs), facilitating ubiquitous AI inference on heterogeneous personal devices. We achieve distributed task offloading via CUDA API remoting. However, beyond raw computation, network constraints emerge as the primary bottleneck: limited bandwidth, high-frequency API invocations, and cross-task contention significantly hinder performance. To address these challenges, we propose Gleam, a novel and network-efficient framework for task-generic GPU sharing across local-area CUDA devices, with three key contributions. First, we reduce bandwidth overhead in CUDA API remoting through automatic model weight caching, and mitigate accumulated latency from frequent API calls by asynchronous execution. Second, we design a runtime task scheduler that dynamically determines API remoting pairs between LAN clients and servers, explicitly accounting for both network conditions and GPU resource contention under parallel workloads. Finally, we introduce dedicated mechanisms to ensure CUDA context consistency across distributed executions. Extensive experiments on heterogeneous NVIDIA GPUs and diverse AI workloads show Gleam consistently outperforms state-of-the-art baselines, achieving 1.4-24.2 times improvements in API remoting efficiency and up to 1.79 times higher system throughput.

Zhihao Xu, Hao Zhong, Zeting Zhou +9

Shanghai Jiao Tong University · Lenovo Information Products (Shenzhen) Ltd · Lenovo (Beijing) Ltd

Heterogeneous Computing Distributed Inference

Jul 21, 2026cs.CR

Integrity of peer-to-peer distributed LLM inference under malicious nodes

Peer-to-peer distributed inference executes a Large Language Model (LLM) on pooled consumer hardware by spreading its layers across many nodes. Every request passes through nodes that are owned and controlled by multiple independent parties. However, in this setting, any party can tamper with the output of its layers to corrupt the end result. Recomputing the forward pass on trusted hardware can catch this, but it introduces additional computational cost. The scientific literature includes several prior integrity-checking approaches, such as known-answer traps for image classifiers and cryptographic commitments. However, these solutions test only the exact correctness and do not account for the ordinary variation that may arise between benign nodes. In this paper, we propose a method that checks the output integrity by measuring the variation in the activations that each node passes to the next. A peer who wants to use the network selects a small set of secret canary inputs whose correct activations are known in advance and mixes them into regular traffic. Because the peers cannot tell a canary from a real query, any tampering node corrupts them as well. The deviation from the known reference then reveals malicious activity: benign nodes exhibit only minor variation from hardware-induced noise, whereas tampered nodes deviate far more. We treat the identification of malicious nodes as a probabilistic test that separates two drift distributions, without relying on a fixed threshold. We study 408 configurations with metrics and success criteria fixed before any experiment ran; the detector reaches AUROC 1.0, correctly ranking the malicious shard above every benign shard on every canary in every configuration.

Mert Cihangiroglu, Antonino Nocera

DCALab, Department of Electrical, Computer and Biomedical Engineering, University of Pavia, Italy

LLM Security Distributed Inference

Jul 17, 2026cs.CV

DiTango: Cost-Effective Parallel Diffusion Generation with Selective Attention State Reuse

Recent advances in AI-generated content have driven widespread adoption of Diffusion Transformers (DiTs) for high-resolution, long-duration content generation. While parallelization techniques accelerate diffusion inference, they face significant scalability challenges due to excessive communication overhead in multi-node environments. We observe that sequence partitions in Context Parallelism (CP) exhibit distinct heterogeneity: spatially proximate partitions contribute more significantly to attention computation results. By mapping this heterogeneous pattern to hierarchical communication topology, we can access high-contribution partitions with reduced communication cost. This insight motivates our novel selective attention state mechanism that strategically balances partial attention computation and historical result reuse across denoising steps. We present DiTango, an efficient parallel framework for DiT generation. DiTango features an anchor-guided state selection planner that optimizes computation-reuse decisions for each partition, complemented by a runtime that orchestrates efficient state-centric operations. This design achieves superior system efficiency while preserving generation quality. Experimental evaluation on popular diffusion models demonstrates that DiTango achieves up to 1.9x end-to-end and 3.2x attention speedup with near-linear scaling in multi-node settings, while maintaining generation quality comparable to state-of-the-art approaches.

Yuyang Chen, Runxin Zhong, Zan Zong +3

Shanghai Jiao Tong University, Shanghai AI Laboratory Shanghai, China · Tsinghua University Beijing, China · University of Science and Technology Beijing Beijing, China +1

Diffusion Transformer Efficient Attention Diffusion Model Caching Distributed Inference Diffusion Model Inference Acceleration

Jul 13, 2026cs.DC

Decomposing Runtime, Kernel, and Quantization Speedups via a Matched FP16 Intermediate: A Hardware-Conditioned Case Study on Four NVIDIA RTX A5000 GPUs

Reported serving speedups from quantized kernels typically bundle the weight format, the kernel, and the inference runtime into one number. We present an attribution study on four NVIDIA RTX A5000 GPUs, 24 GiB each, on a single host with NVLink-bridged pairs. A matched intermediate stack that keeps the faster runtime without the quantized kernel splits the full speedup into a runtime part and a kernel and quantization part. Under matched greedy decoding the full stack reaches

2.58\times

end to end, with the runtime change accounting for about two thirds of that gain on a logarithmic scale; across three similar model families the kernel and quantization part moves by at most 1.5%. Sharding one instance across all four cards falls well below doubling: a profiler trace attributes about 80% of the per token shortfall to coordination, and an NVLink versus PCIe control on the same hardware shows similar realized bandwidth on both links, pointing away from link bandwidth as the cause. Whether to run one sharded instance or several independent ones depends on the workload and the model, with the ranking reversing on the larger model: the smaller model splits between sharding and multiple instances by workload, while the larger model favors two paired instances on every workload. Quantization extends sustainable concurrent users roughly four times past a reproducible half precision memory cliff. Differences in sampling mode and prompt pool between the two stacks are documented as threats to validity.

Weijia Han, Lisha Qu

University of Washington

LLM Quantization GPU Kernel Optimization LLM Inference Scheduling LLM Inference Acceleration Distributed Inference

Jul 11, 2026cs.DC

Stateful Worlds, Stateless Elasticity: Exact-State Serving for Interactive World Models

A persistent interactive world model keeps its running state resident on the GPU that serves it: a multi-gigabyte attention cache, almost all of it rewritten at every generation step. That state cannot be recomputed in interactive time or approximated without changing the world, so a live session pins its device. The pin is a scheduling problem. WorldMove moves a live session under one guarantee: the destination is bit-identical to the source, or nothing is installed. It relocates the cache in 18.8 ms same-node, 101x faster than save/load. It holds a checksum-verified 92.1-94.8 Gb/s on a 100 Gb fabric. At that rate the cache fits inside one interactive block. Migrating an actively generating session, it converges at a block boundary and the destination continues the world bit for bit. An admissibility condition decides each move. The move must complete inside the readout horizon, over bandwidth that covers the state plus its dirty rate. Lifted to a fleet schedulability test, it governed a consolidation loop that executed 48 of 48 migrations bit-identical across two providers. Two constraints are structural. Bit-exactness survives only inside a controlled configuration of one GPU architecture, so moving the state is the only way to preserve it exactly in interactive time. Verification cannot hide inside the wire on this fabric. Receive-path checksums stall the transport at protocol timescales under fan-in, and unscheduled incast silently collapses a receiver while every delivered byte stays correct. An incast-aware admission controller holds zero misses to 1.4x offered load and sheds overload as rejects. A lossless GPU codec widens the admission gate to fabrics raw motion cannot use. We exercise the serving loop and the mover separately, each end to end. Their composition on one fabric is unbuilt. Exact-state elasticity is a joint scheduling problem over transport and verification.

Jin Li, Jiawei Chen

Harvard University · Harvard University Cambridge, MA, USA · Independent Researcher

Distributed Inference Online Resource Allocation Interactive World Models

Jul 9, 2026cs.RO

D-CLIPSE: Distributed Consensus-based Localization with Passive Listening on Shared State Exchange

Multi-robot localization that is accurate and consistent is imperative for downstream tasks such as planning and control. Centralized filtering approaches optimally fuse all available sensor measurements of the team. However, a centralized solution is rarely implementable due to hardware, communication, and computational constraints. Distributed approaches deploy a filter on each robot to estimate their own state and neighbours' states using inter-robot communication. This paper proposes a consistent, communication-efficient, and consensus-based distributed filtering framework that shares both preintegrated odometry and relevant shared states among communicating robots. The proposed method is validated in simulated and experimental scenarios, showing near centralized performance in accuracy, and especially in consistency, compared to the current state-of-the-art decentralized approach.

Kyle Biron-Gricken, James Richard Forbes

Cooperative Localization Robot State Estimation Distributed Inference

Jul 8, 2026cs.DC

CTA-Pipelining: A Latency-Oriented Spatial Scaling Method for Multi-GPU Systems

The evolution of compute infrastructure has transformed multi-GPU systems into tightly integrated shared-memory structures. However, current software still mostly treats these coherent interconnects simply as high-speed networks. Simultaneously, the demand for serving Large Language Models under latency constraints has shifted GPU workload optimization from being throughput-driven to latency-bound, necessitating latency-oriented scaling methods beyond Tensor Parallelism (TP). Thus, we introduce CTA-pipelining, an execution paradigm designed to exploit shared-memory multi-GPU systems. As a latency-oriented spatial scaling technique, CTA-pipelining leverages dependencies at the Cooperative Thread Array level, enabling concurrent execution of dependent kernels across GPUs. We demonstrate its capability using CUTLASS, cuBLAS, and NCCL libraries on 8-GPU H200 and B200 systems. Results show on 2-layer GEMM, representing the MLP operation, CTA-pipelining reduces latency by up to 31.8% compared to micro-batching, and 29.6% compared to TP. It can also be combined with TP as an orthogonal scaling dimension to further push the latency boundary.

Tingkai Liu, Muralidhar Andoorveedu, Sanjoy Das +2

University of Illinois Urbana-Champaign · NVIDIA Corporation

GPU Kernel Optimization High-Performance Computing LLM Inference Acceleration Pipeline Parallelism Distributed Inference

Jul 3, 2026cs.LG

Design-CP: Context Parallelism for Design of Protein Nanoparticles

Many all-atom generative protein models can in principle design large multimeric complexes by jointly modelling all chains, but their quadratic token- and atom-pair representations quickly exceed single-GPU memory as the number of chains and residues modelled grows. We introduce Design-CP, two context-parallel (CP) inference strategies for RFdiffusion 3 (1D row-sharding and 2D grid sharding with ring attention) that distribute the quadratic activations across a multi-GPU mesh while preserving pretrained weights. We characterise their scaling when sampling icosahedral assemblies, showing that the maximum feasible asymmetric subunit (ASU) size grows with the expected square-root trend in GPU count and that 2D sharding achieves better wall-clock scaling. Moreover, we show how strong point-group symmetry constraints make CP usable out of the box for end-to-end, all-atom design of icosahedral nanoparticles, yielding favourable in silico structural and interface metrics. Finally, we demonstrate octahedral nanoparticle design on a small cluster of workstation-grade 16GB GPUs, illustrating how Design-CP can be a practical path towards democratising large-assembly protein design.

Lorenzo Tarricone, Helen E. Eisenach, Aiko Muraishi +1

Department of Statistics, University of Oxford, Oxford, UK · Ellison Institute of Technology Oxford, Oxford, UK · Institute for Protein Design, University of Washington, Seattle, USA +2

GPU Acceleration High-Performance Computing Protein Design Distributed Inference

Jun 30, 2026cs.DC

From Tensor Buffer to Distributed Memory Hierarchy: A Survey of KV Cache Management for LLM Serving

The key-value (KV) cache has become a first-order memory object in LLM serving rather than a temporary per-request tensor. This survey classifies more than thirty KV-management systems and frameworks using four axes: locality, lifetime, ownership, and substrate. The axes reveal five architectural archetypes -- local-paged, disaggregated-pipeline, shared-store, memory-pool, and hybrid-tier. Once workload and hardware are fixed, ownership accounts for much of the remaining design variance among distributed systems. The survey also audits current evaluations and identifies seven missing KV-specific measurements, linking them to open problems in fault tolerance, isolation, tiered eviction, speculative decoding, MoE serving, and shared-cache semantics.

Jie Li, Tongyang Wang, Yong Chen

Texas Tech University, Department of Computer Science, Lubbock, Texas, USA

LLM Inference Efficiency LLM Serving KV Caching KV-Cache Management Distributed Inference

Jun 23, 2026cs.AI

Decentralised AI Training and Inference with BlockTrain

Frontier AI training is increasingly shaped by access to dense, centrally controlled accelerator clusters. This creates a structural advantage for hyperscalers and large centralized laboratories, and makes open or independent AI efforts depend on scarce capital, privileged infrastructure, and data-center geography. We present Spheroid BlockTrain, a decentralized training protocol in which a model is partitioned into independently trainable blocks, each optimized on a local objective derived from the same global target and composed at inference into one model. On byte-level WikiText, BlockTrain reaches cross entropy 1.359 (perplexity 3.89), within about 0.04 CE of a same-setup end-to-end Transformer reference, while each active worker trains only one block and avoids full-model optimizer state. A shared six-worker block training run reaches CE 1.385 by averaging same-block updates into one assembled model. HTTP/TCP transport experiments move real serialized checkpoints and updates, including a public-IP three-host run that improves CE from 5.580 to 1.811 while moving 15.22 GB. For inference, the current BlockTrain path uses one block-stack traversal per full output and serves over direct TCP across three public-network GPU hosts up to a 75.80B-parameter logical fp16 shape, outperforming a matched plain-autoregressive TCP pipeline baseline because it emits a full sequence per WAN pipeline traversal rather than one token per traversal.

Peter Toth, Dan Oprisa

Spheroid Labs∗

Efficient Transformer Inference Decentralized Learning Distributed Training Distributed Inference

Jun 22, 2026q-bio.GN

Privacy-preserving federated tensor decomposition of single-cell immune data: recovering multicellular programs across institutions

Tensor decomposition of donor

\times

cell-type

\times

gene single-cell data recovers \emph{multicellular programs}: coordinated axes of inter-individual transcriptional variation that span cell types and stratify disease. Yet immune single-cell atlases are increasingly multi-institution, multi-ancestry, and governed, so patient cells often cannot be pooled. We present a federated estimator: each site computes a local program subspace, and a coordinator merges these by stacked SVD under federated global-mean centering, provably equivalent (up to truncation) to the centralised decomposition. This centering makes the merge robust to site-label confounding (program AUC

0.957

vs.\

0.861

for naive per-site centering). Only program subspaces leave a site, and aggregation is compatible with secure aggregation. On a 261-donor systemic lupus erythematosus atlas it recovers the canonical interferon program (ISG enrichment AUC

0.998

; case--control separation

0.958

; bootstrap

Δ\text{AUC}=-0.000

, 95% CI

[-0.004,+0.012]

vs.\ centralised), across institution-scale and multi-ancestry partitions, and across three \emph{real} COVID-19 sites (subspace correlation

0.989

). It recovers the program when \emph{no site observes all cell types} (correlation

1.000

, exact by construction), which fixed-feature federated PCA cannot. On an interstitial-lung-disease atlas the recovered program predicts disease better than the best single cell type (AUC

0.96

vs.\

0.91

; gap 95% CI excludes zero) and the advantage survives federation; a liver cohort is consistent (

p=0.005

). Membership-inference shows secure aggregation cuts attack AUC from

0.91

to

0.61

. The method enables cross-institution, cross-ancestry recovery of multicellular immune programs without sharing cells.

Axel Faes, Stephanie M. van den Berg, Maryam Amir Haeri

Data Science Institute, Hasselt University, Belgium · University of Twente, The Netherlands · Learning, Data Analytics and Technology, University of Twente, The Netherlands

Federated Learning Aggregation Tensor Decomposition Distributed Inference Single-Cell Transcriptomics Federated Learning

Jun 19, 2026cs.DC

BatchGen: An Architecture for Scalable and Efficient Batch Inference

Batch inference has become a central mode of AI computation, yet existing inference engines still rely on execution models designed for interactive serving. When scaled to millions of sequences, batch workloads reveal two fundamental requirements: the ability to handle extreme inter- and intra-sequence load variation that emerges only at runtime, and the ability to sustain high utilization across large fleets of GPUs. Existing systems fail to meet these requirements, losing substantial fractions of achievable throughput. We introduce a new architectural foundation for batch inference: the sequence coroutine compute model, which represents each sequence as a fine-grained, event-driven coroutine. This model exposes expressive primitives that allow the runtime to reorganize work dynamically, enabling larger expert-level batches, mitigating stragglers, reallocating work across devices, and maintaining utilization even on cost-effective or memory-constrained GPUs. Building on this abstraction, we implement BatchGen, a production-ready system that uses the coroutine model at cluster scale. On a 128-GPU cluster, BatchGen reduces batch completion time by up to

2.3\times

, and on memory-constrained accelerators it outperforms the strongest offloading baseline by up to

9.6\times

. We will open-source BatchGen at https://github.com/batchgen-project/batchgen

Tairan Xu, Leyang Xue, Zhan Lu +7

University of Edinburgh · Tencent

Efficient Neural Network Inference Efficient Inference Distributed Inference

Jun 17, 2026cs.MA

Mesh Inference: A Formal Model of Collective Inference Without a Center

We present a formal model of mesh inference: how a population of independent agents, each holding private state and exchanging only admitted, typed observations, derives a conclusion none of them holds alone, with no central coordinator and no agent exposed. No agent shares weights, gradients, or hidden state, and the agents may span different teams, networks, and organizations. Motivated by the observation that asking a model is energy-minimizing inference, we model the mesh as a coupled free energy that each agent relaxes locally. We show that a single admission/emission policy governs three properties. First, mesh inference converges to a unique answer for any admission, symmetric or not, because the coupling is always an M-matrix. Second, it is identification-complete: it derives the centralized optimum exactly when the contributing views are carrier-connected. Third, it is observation-only: no node transmits its internals, and confidentiality is the dual of identification. Content-addressed lineage is the only global side-channel. In the linear-Gaussian regime every derived answer is determined, hence equal to the centralized optimum, at O(diam^2) latency, the measured price of removing the center. One such derivation is one turn of a center-free learning loop, which we formalize as architecture rather than prove. The open problem we state is when asking improves the collective rather than corrupting it: whether the non-linear closure derives an upgraded answer or a confident error. To our knowledge, this is the first formal characterization of when a center-free, observation-only mesh recovers the centralized optimum.

Hongwei Xu

Privacy-Preserving ML Multi-Agent Reasoning Distributed Inference

Jun 16, 2026cs.DC

AoiZora: Topology-Aware Auto-Parallel Optimization for Inference of Diffusion Transformers

Video diffusion has quickly grown into a key generative serving workload, yet producing each clip demands many denoising iterations over large spatio-temporal latents, which puts low-latency inference out of reach on a single device. A denoising step is therefore typically distributed across multiple accelerators, and TPU sub-slices have become an attractive and practical fabric for doing so. Current auto-parallel systems, however, search almost exclusively over logical device meshes and disregard how a chosen sharding is actually laid out on the physical TPU interconnect -- an oversight that leaves large, topology-dependent performance on the table. We address this gap with AoiZora, a compiler-mediated topology planner built for low-latency video diffusion inference on TPU sub-slices. Its guiding principle is to reconnect logical sharding with physical placement by drawing on different points in the compilation flow: AoiZora first eliminates weak sharding candidates from inexpensive pre-compilation IRs, then compiles only the ones that survive and orders their physical placements using compiled HLO together with a topology-aware communication model. The winning plan is realized along the ordinary compiler path, leaving model code, compiler lowering, collective kernels, and network routing entirely intact. On TPU v5e sub-slices, AoiZora reduces Wan 2.1 one-step denoising latency by as much as 1.42x relative to existing solutions.

Kaijian Wang, Yuanyuan Xu, Fanjiang Ye +5

Rice University · Independent Researcher · Google

AI Accelerator Inference Video Diffusion Models Diffusion Transformer Distributed Inference Diffusion Model Inference Acceleration

Jun 13, 2026cs.LG

Director: Accelerating Distributed MoE Serving via Online Proactive Expert Placement

Expert parallelism has become the prevailing paradigm to serve Mixture-of-Experts (MoE) models. Its efficiency depends on the communication and computation latencies of the GPUs, which are linked to the placement of experts in the GPUs. Existing works for optimizing expert placement focus on leveraging past requests' expert activation patterns. However, they demonstrate deficiencies facing diverse and rapidly changing request patterns, calling for an online, proactive approach. Implementing such an approach requires addressing several challenges: the uncertainty associated with incoming requests' expert activation, the cost of expert migration, and the NP-hard complexity in optimization. Therefore, we present Director, a new distributed MoE serving system that minimizes end-to-end latency via prediction-driven, online expert placement. Director uses either a lightweight cascaded predictor or a low-bit quantized replica for expert activation patterns of incoming requests. An online migration module then enacts the changes with near-zero downtime by executing migrations in compute-bound phases, keeping disruption bounded. At its core, a relaxation-based expert placement optimizer operates under capacity constraints, runs in polynomial time, and achieves a

(1+ε)

approximation ratio. Finally, we implement a prototype and demonstrate, through extensive experiments, a reduction in end-to-end latency of

11\sim55\%

for popular MoE models (e.g., Mistral, DeepSeek and Qwen) compared to existing work.

Qianli Liu, Kaibin Guo, Zicong Hong +5

Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong · School of Software Engineering, Sun Yat-Sen University, China · School of Cyber Science and Engineering, Xi’an Jiaotong University, China +1

Expert Parallelism LLM Serving Mixture-of-Experts Inference LLM Inference Acceleration Mixture-of-Experts Models Distributed Inference

Jun 11, 2026cs.AR

The Price of Anarchy in Disaggregated Inference

Disaggregated inference architectures physically separate prefill and decode phases onto distinct GPU pools, creating competing "agents" that share a fixed hardware budget. We provide, to our knowledge, the first formal game-theoretic analysis of this architecture, using NVIDIA Dynamo as a concrete case study. We model disaggregated serving as three coupled games: a two-player resource game between prefill and decode pools, a selfish caching game over the hierarchical KV cache, and a congestion game with positive externalities for request routing. We empirically validate the latter two; the P/D resource game is treated analytically (Section 9.2). We characterize how GPU saturation induces regime transitions that shift the game's payoff structure: below saturation, selfish behavior has bounded Price of Anarchy (PoA); at saturation, superlinear latency and cache externalities drive our empirical estimator PoA-hat (defined in Section 6.4) upward. Based on this analysis, we design an adaptive controller that detects saturation transitions in real time and adjusts routing parameters accordingly, shifting from cache-affinity exploitation to load-balanced congestion avoidance. We instantiate our framework on a 3-node NVIDIA B200 cluster running Dynamo with two models, Nemotron-4-340B (TP=8, full-node workers with cross-InfiniBand KV transfers) and Llama-3.1-70B (TP=4), and find the same three-regime PoA-hat structure with the same first post-knee grid point (C=128) on both models. Adaptive routing shifts each model to a better operating point. Our strongest result is on the 70B 1P/5D topology, where PoA-hat drops 3.1x (66.4 to 21.5) in the saturated phase at a 13% throughput cost. On the 70B 1P/2D, PoA-hat drops 2.2x and TTFT P99 drops 7.6x (see Section 8.5).

Athos Georgiou

NCA

LLM Inference Efficiency Game Theory Distributed Inference

Jun 10, 2026cs.LG

M*: A Modular, Extensible, Serving System for Multimodal Models

We are entering a new era of composite model architectures that integrate diverse components such as vision encoders, language backbones, diffusion and flow heads, audio codecs, action generators, and world-model predictors. Such architectures underpin a broad class of multimodal models, including unified multimodal models, omni models, speech-language models, vision-language-action policies, and world models. However, existing model serving frameworks were built on narrow assumptions about model structure, making them ill-suited to accommodate this new architectural diversity. Here we present M*, a universal serving system for efficient serving of composite AI models. M* represents models as dataflow graphs, processing requests spanning diverse modalities and tasks as traversals over these graphs. The core insight is a modular abstraction that supports arbitrary composition of model components, flexible placement onto a physical cluster, and model-agnostic optimizations within a distributed runtime. We call this abstraction the Walk Graph and show how it can concisely capture composite models from a broad range of families. We instantiate M* on representative models and find that it achieves, on average, 20% lower end-to-end latency than vLLM-Omni for text-to-image workloads on BAGEL, while delivering up to 2.9x lower real-time factor and 2.7x higher throughput for text-to-speech workloads on Qwen3-Omni. M* also outperforms the V-JEPA 2-AC rollout baseline for robotic planning by up to 12.5x. Thus, our work paves the road towards more efficient serving of complex models with minimal developer effort.

Atindra Jha, Naomi Sagan, Keisuke Kamahori +9

Stanford University · University of Washington · Carnegie Mellon University

Efficient Multimodal Inference Distributed Inference

Jun 3, 2026cs.DC

Multi-SPIN: Multi-Access Speculative Inference for Cooperative Token Generation at the Edge

Speculative inference (SPIN) was originally developed as an efficient architecture to accelerate Large Language Models (LLMs). In this work, we propose its distributed deployment to enable cooperative token generation in a multiuser edge system; its advantage is to effectively balance computational loads between resource-constrained devices and servers. The resulting architecture, termed Multi-access SPIN (Multi-SPIN), utilizes on-device small language models to generate and upload candidate token drafts, while an edge server operates the LLM to verify them in parallel batches. Given the severe heterogeneity in users' computation and communication capabilities, the draft length emerges as a critical control variable that influences node-level computation loads and multi-access latency, thereby governing the sum token goodput. Consequently, considering frequency-division multiple access, we investigate the problem of multi-access draft control, a joint optimization of draft-length control and bandwidth allocation to maximize sum token goodput. We examine two cases: (1) homogeneous draft lengths across users to facilitate server-side batching, and (2) heterogeneous draft lengths to introduce a new dimension for goodput enhancement. By developing decomposition methods, we reduce these complex optimizations into tractable sub-problems, which allow efficient draft control algorithms to be derived in closed form. Our analysis shows that the optimal bandwidth allocation compensates users with weaker computation-and-communication capabilities in the homogeneous case due to the batching synchronization requirements, whereas its heterogeneous-case counterpart rewards users with higher acceptance rates by relaxing such requirements. Experiments using Llama-2 and Qwen3.5 model pairs across diverse tasks demonstrate that Multi-SPIN improves goodput by up to 88% over heterogeneity-agnostic baselines.

Haotian Zheng, Zhanwei Wang, Mingyao Cui +3

Department of Electrical and Computer Engineering, The University of Hong Kong, Hong Kong SAR, China

On-Device Language Model Inference Edge Computing Speculative Decoding LLM Inference Acceleration Wireless Resource Allocation Distributed Inference