Low-Latency Systems

Modern smart vehicles leverage multimodal sensors, ranging from high-bandwidth vision systems to low-rate physiological monitors, to provide personalized in-cabin services. However, integrating high-fidelity multimodal fusion with collaborative training is often hindered by the heterogeneous and time-varying Quality of Service (QoS) constraints of vehicular networks. Standard Federated Learning (FL) approaches enforce rigid synchronous rounds that fail to account for these resource asymmetries, leading to safety-critical timing violations and energy exhaustion. In this paper, we propose FedQoS, a novel asynchronous, event-triggered FL framework that decouples local computation from global communication via a two-phase gating mechanism. First, we introduce a resource-aware training gate that initializes local learning only when sensing buffers and energy reserves meet safety thresholds, preventing ML tasks from compromising core vehicle mobility. Second, a QoS-aware transmission policy gates uplink updates based on an efficiency score that balances model novelty against instantaneous latency and energy costs. Locally, clients optimize an objective featuring a staleness-aware proximal term that dynamically adjusts the global anchor strength based on update age. Extensive experiments on multimodal vehicular datasets demonstrate that FedQoS achieves competitive personalized accuracy with only marginal performance loss compared to FedAvg, while substantially reducing QoS violations, cutting communication overhead by 76.7%, and lowering latency cost by 26.0%, demonstrating a highly favorable accuracy and efficiency balance for real-world vehicular deployments.

Modern networks must support changing topologies, configurations, and performance objectives, motivating fast and reliable performance estimation. Network digital twins (NDTs) enable what-if analysis for performance estimation in such network scenarios, however, existing machine learning-based NDT approaches often rely on entire topology representations, which are inherently monolithic and lack reusability under topological or traffic changes in the network. This paper introduces a composable NDT approach that decomposes networks into subgraphs represented by reusable unit twins that capture subgraph structure, configuration and traffic behaviours. A lightweight composer aggregates unit twin combinations to create NDTs that predict per-route end-to-end latency through an overall topology. Evaluation across controlled synthetic topologies and diverse traffic scenarios, real-world Topology Zoo topologies, and a public NDT challenge dataset demonstrates that the composable NDTs achieve high in-distribution accuracy while remaining stable under out-of-distribution scenarios. Comparison with monolithic full topology NDTs demonstrates that our composable approach achieves reusability, while achieving comparable or superior accuracy.

Streaming automatic speech recognition (ASR) must be judged jointly on what it transcribes and on how quickly it commits each word. Delayed streams modeling (DSM) has become the dominant paradigm for streaming large audio-language models, exposing a structural delay $τ$ that bounds the decoder's lookahead. We show that $τ$ is a poor proxy for user-perceived latency, and that the alignment-based supervision of DSM leaves latency on the table: the same forced-aligned transcript is used at every $τ$, forcing the model to withhold words it could already commit. We introduce AWED, a word-level emission-delay metric defined relative to the acoustic end of each word, and post-train a DSM recognizer with GRPO under a reward that scores transcription accuracy and measured delay jointly. Trained at a single operating point ($τ=6$ frames), our model dominates both its supervised fine-tuning initialization and the Voxtral Realtime backbone across all evaluated lookahead budgets: it cuts WER by 30.8% relative at an 80,ms structural delay, and by 5.7% relative at 480,ms while lowering median AWED from 1.17,s to 1.04,s. Latency-rewarded post-training thus advances the accuracy--latency Pareto frontier of streaming ASR without architectural change.

LLM serving is typically offered as a shared, multi-tenant service, where high-demand workloads from one client can cause latency SLO violations for others. Existing solutions for performance isolation equalize client throughput in the long run, for example through queueing and batching fairness. However, these approaches do not provide latency isolation guarantees; as a result, well-behaved clients can still experience significant degradation to their token-level latencies. In this paper, we present FairInference, which provides the novel δ-token fairness guarantee: for a well-behaved client, if a token is generated in d time units in isolation, it will be generated within d + δ time units in multi-tenant execution, providing strong latency isolation guarantees for LLM serving. To achieve this, FairInference addresses a key challenge of LLM serving: bounding delays from sharing GPU resources without support for fine-grained scheduling or resource allocation. In FairInference, the scheduler enforces per-token deadlines, while bounding the delays from GPU compute sharing and accounting for the additional delays introduced by the shared KV caching in GPU memory. We show that FairInference effectively bounds token-level latency spikes for well-behaved clients and improves overall throughput compared to state-of-the-art LLM serving systems.

Warm-start transfer can make algorithmic tasks generalize rapidly, yet it is unclear which model components provide the gain and whether that gain remains stable under continued optimization. We study cross-operator transfer on modular arithmetic and separate efficacy (early velocity) from stability (post-reach drawdown). In a scale-matched 108-run battery across 12 seed blocks (96-run 2^3 factorial plus 12-run scale control), transferring internal attention/MLP weights (B) alongside token embeddings and readout (E+U) improves early accuracy by 5.46 pp (Holm p=0.0039) and cuts confirmation latency by 558 steps (Holm p=0.0088). While readout plus internal-block transfer satisfies the pre-specified +/-500-step latency equivalence criterion in 1-layer models (TOST p=0.0011, though Full is faster in 11/12 paired seeds), a prospective 2-layer replication confirms the internal-block advantage (12/12 seeds, +704.67 integral units, p=4.88x10^-4) while revealing an architectural boundary: omitting donor embeddings falls 4475.6 units below Full, outside the +/-250-unit margin. Continued target training frequently triggers severe post-grokking relapse. Freezing transferred representation carriers (E, U) nearly eliminates offline relapse (19.40% -> 0.07%, Holm p=0.005859). Online validation-triggered gating slashes True Max Drawdown from 22.06% to 0.60% on 2a+b (p=0.000488), with prospective confirmations extending protection across affine, nonlinear quadratic, and 2-layer targets (10.94-23.47 pp reductions), distinguishing continual stabilization from static early stopping. In non-abelian S_5, unshielded transfer surges transiently (95.4% peak), but a prospective shielding cohort yields no confirmed benefit (+0.15 +/- 1.14 pp). These results establish a component-level dissociation between transfer acceleration and trajectory stability, and expose the empirical boundaries of parameter shielding.

LLM inference optimizations report speedups on different models, GPUs, prompts, and quality metrics, making them hard to compare or combine. We build a cost, quality, and latency Pareto atlas to identify the best configurations for different deployment constraints. Since exhaustive testing is impractical, we measure 54 configurations of Qwen2.5-7B-Instruct running on vLLM 0.12 across L4, A100, and H100 GPUs and use these anchors to calibrate a simulator. It reproduces measurements at anchored batch sizes, with cross campaign drift below 1.5 percent. A separate quality evaluation tests FP16, AWQ 4bit, FP8 weights, and FP8 KV cache on 200 GSM8K questions with five examples per prompt. Sparse attention is evaluated only in simulation. On the calibrated grid, 18 of 36 configurations reach the Pareto frontier. Combined methods reach it more often than individual methods, with 9 of 15 combinations versus 9 of 21 single methods. Quality testing changes the winners. AWQ 4bit reduces per token latency to 0.34 times baseline on L4 but loses 5.9 percent of strict GSM8K accuracy, narrowly missing the 95 percent quality floor within sampling uncertainty. Flexible answer extraction matches FP16 accuracy, suggesting the loss comes from formatting rather than arithmetic. FP8 weights retain 99.4 percent of baseline accuracy at 0.61 to 0.65 times baseline latency across all three GPUs and appear in three of four regime winners. A naive FP8 KV cache maintains normal throughput but answers none of the 200 questions correctly, showing why speed alone is insufficient. Under two prompt designs, n gram speculative decoding measures at 0.90 to 0.98 times baseline and adds no benefit on this stack. The best choice depends on the constraint and GPU: H100 wins for tight latency, while A100 wins for throughput and low cost at 0.106 dollars per million tokens.

One-to-many machine translation (MT) is computationally expensive for autoregressive (AR) systems, which suffer from linear latency scaling with both sequence length and the number of target languages. We explore how diffusion can enable multilingual translation with a discrete diffusion framework that refines all target languages in parallel, achieving sublinear latency scaling with the number of targets, and supports deployment as a single unified model to replace multiple independent systems. Conditioned on a continuous semantic anchor rather than source tokens, our framework supports zero-shot transfer to unseen source languages without retraining, maintaining approximately $75\%$ of its supervised translation quality on zero-shot sources. We investigate the quality-latency frontier and find that with accelerated sampling, it achieves comparable supervised quality to AR baselines with a $2 \times$ speedup and $11.9\%$ better zero-shot BLEU. These results highlight the potential of joint multi-stream diffusion as a practical and flexible alternative for efficient one-to-many translation.

Emerging physical AI systems require low-latency, task-oriented video communication over unreliable channels. We propose a semantic-aware multi-level neural video coding method for robust low-latency video transmission over unreliable channels that are abstracted as multi-level packet erasure channels. Built upon the real-time DCVC-RT neural video codec, the proposed framework introduces a semantic- and feature-aware coding strategy that partitions encoded representations into packets carrying different levels of semantic and latent-feature importance and assigns these packets to different streams, each associated with a priority level when transmitted over unreliable communication channels. We also developed an error-resilient entropy model that removes inter-packet dependencies, allowing each packet to be decoded independently under packet losses. The complete system is trained end-to-end over the abstracted multi-level packet erasure channels, enabling learning of channel-aware representations together with importance-aware packet assignment while facilitating the network for differentiated packet prioritization. Experiments show that the proposed framework significantly improves robustness over baseline DCVC-RT under packet erasures, achieving graceful degradation in less important regions while better preserving task-relevant visual content.

Slip detection is fundamental to dexterous manipulation, yet existing systems often lack precise characterization of detection latency and cross-platform generalization. We present SlipSense, a multimodal tactile slip-detection framework built on TacV5, a compact sensor integrating a $32 \times 32$ piezoresistive array operating at 240 Hz and a 3-axis MEMS accelerometer operating at 8 kHz. The piezoresistive array captures spatial pressure distributions, while the accelerometer captures friction-induced vibrations, providing complementary slip cues. The framework performs modality-specific encoding, intra-sensor fusion, and cross-modal attention with causal temporal prediction at 240 Hz. Experiments on a dataset of 1.4 million frames spanning 37 objects demonstrate the complementarity of the two modalities. SlipSense achieves 96.7% Macro F1 with a false-positive rate below 1.6%, detecting 76% of slip events within 23.1 ms. When trained solely on UMI data, SlipSense generalizes zero-shot to a Tesollo dexterous hand, transferring across unseen objects, distinct sensor units, and robotic platforms without retraining.

Tabular foundation models deliver strong zero-training predictive performance via in-context learning, but their high inference latency makes them impractical as hot-path decision backends in interactive agentic loops. We distill a TabPFN teacher into a compact feed-forward student across a business-decision simulation on UCI Adult and five OpenML benchmarks: the classification head compresses 53.2M parameters to 8,546 (6,220x); the deployed two-head loan pipeline compresses 111.4M parameters to 17,059 (6,532x). The student retains 95.4-100.5% accuracy and 96.8-100.0% AUC, with the lowest accuracy retention on credit-g at 95.4%; an alpha = 0 hard-label control shows that the teacher's soft targets provide a 2.1-7.0 AUC point gain.

The surge of large language model (LLM) applications on personal devices imposes massive, bursty workloads on cloud serving infrastructure. While prefill-decode disaggregation improves throughput and scalability, memory-bound decode instances often suffer from persistent load imbalance, as output lengths are unknown when requests arrive at the cloud. To address this, we propose MAPS, a Memory-Aware Predictive Scheduling framework tailored for disaggregated LLM serving. MAPS performs device-assisted speculative output length prediction overlapped with cloud-side prefilling, incurring negligible latency overhead. To handle generation uncertainty, MAPS applies uncertainty-aware calibration to derive output-length upper bounds with target coverage, enabling safe scheduling decisions. Building on these bounds, MAPS employs a hierarchical global-local scheduling strategy to mitigate inter-decoder queue buildup and intra-decoder head-of-line blocking. Extensive experiments on two real-world workloads and two LLMs show that MAPS significantly outperforms three state-of-the-art systems, reducing average end-to-end latency by 42.6 and tail latency by up to 84.8.

Real-time music source separation is validated on desktop CPUs and GPUs. Does any published system fit the embedded audio hardware it targets? On a commercial audio DSP (2 MB SRAM, 2.07 GMAC/s measured), none does, and the constraints eliminate different models: memory rules out the 16-51 M parameter TasNet/X-UMX family, per-frame compute rules out RT-STT, needing 5.5x the available MAC rate. Parameter count predicts neither: weight reuse spans 1x to 345x. We then build one that fits. Training on continuous rather than block-padded convolution context proves essential: a model scoring 3.93 dB block-wise otherwise collapses to silence within 2 s frame-by-frame. A gated complex FIR deep filter adds a latency knob, gaining 0.38 dB even when strictly causal. It reaches 4.70 dB cSDR on MUSDB18-HQ and runs in 10.43 ms of an 11.6 ms hop, 0.5-0.7 dB behind systems that do not fit.

Foundation models, including large language models, vision-language models, and time-series foundation models, are increasingly deployed on embedded and edge platforms for CPS and IoT applications, where energy, latency, and memory are as critical as task accuracy. Existing benchmarking tools evaluate model capability in isolation, reporting accuracy assuming sufficient compute, while hardware profiling tools remain platform-specific and mutually incompatible. As a result, users lack a unified workflow for making deployment decisions across heterogeneous devices. We present HoliBench, a modular benchmarking and deployment toolkit that jointly characterizes accuracy, latency, and energy across platforms from single-board computers to GPU servers. Its platform abstraction layer calibrates cross-device measurement, and the toolkit supports multiple model modalities, inference engines, concurrencies, and existing evaluation harnesses. An interactive interface exposes constraint-aware configuration selection over a design space that is profiled once and reused across studies. Using HoliBench, we characterize 20 models across 7 device types, 3 quantization levels, 8 inference backends, and over 30 tasks, surfacing tradeoffs that existing tools miss: quantization reduces latency only on hardware with low-precision support, accuracy gains show diminishing returns relative to energy, and for autoregressive workloads, average inference power is approximately constant across output lengths. We further find that single-model profiles compose under sequential co-resident execution. In a multi-model CPS deployment, standalone profiles predict combined-pipeline latency and power within 1.2% and 2.5%, enabling deployment exploration without exhaustively profiling every pipeline configuration. We release HoliBench as open-source infrastructure for deployment-aware evaluation of foundation models.

Simulated students generated by computational models provide a practical way to evaluate tutoring strategies and pedagogical approaches used by human and AI tutors. However, such simulations should account for disengaged behaviors, including gaming the system, wheel-spinning, and off-task behavior, because tutors may need different responses for different learner states. We present Disengagement-Aware Student Simulators (DAS2), a reproducible pre-deployment protocol that models five learner-engagement states: engaged, gaming, wheel-spinning, off-task, and mixed, and evaluates AI tutor performance across these states. Using ASSISTments09, two coders independently labeled 100 sampled tutoring sessions based on anonymized interaction-log summaries. They achieved 84% agreement (Cohen's kappa = 0.78), and among agreed cases, human consensus labels matched DAS2 rule-based labels in 81% of cases (kappa = 0.75). Conditioning simulations on intended learner states reduced the correctness-rate gap between simulated and authentic sessions from 0.54 to 0.20 for gaming and from 0.51 to 0.18 for wheel-spinning. Fine-tuned Qwen2.5-7B better matched authentic response-time distributions, while prompt-only GPT-4o generated more distinguishable learner states. Evaluation of five AI tutors from the Claude, Llama, Gemini, Qwen, and GPT families showed that relative rankings remained stable across learner states and interaction lengths, while absolute performance varied, revealing state-specific differences in tutor support. Human validation further showed that automated tutor evaluation does not fully align with human judgment. DAS2 provides a pre-deployment framework for evaluating how AI tutors respond to diverse learner-engagement states before deployment.

We model total latency as the convolution of preprocessing, inference, and postprocessing distributions under a simplifying independence approximation, with selected stage parameters expressed as functions of source-image resolution. Under this assumption, the probability density of the total latency is the convolution of the stage-wise densities, and its cumulative distribution function (CDF) provides the distribution of end-to-end detection time. Each stage is modeled by a parametric distribution (e.g., Exponential, Erlang, Normal, Gamma), with parameters expressed as functions of the source-image resolution. Experiments with YOLOv11n on NVIDIA Jetson Orin NX using COCO2017 images across multiple resolutions assess the proposed models against fixed-parameter baselines using Kolmogorov Smirnov, Anderson Darling, and Cram'er von Mises statistics. The results indicate that resolution-aware parameterization can improve distributional approximation in the measured setting, particularly for the more flexible Normal and Gamma models, while the quality of fit remains distribution dependent. Our contribution is a theoretically grounded and lightweight formulation for studying resolution-dependent latency distributions in a measured object detection pipeline.

Vision-language-action (VLA) deployment can reduce inference latency while changing closed-loop task behavior. We evaluate HuggingFaceVLA/smolvla_libero on an RTX 2060 (6 GB) in LIBERO Spatial and Object (MuJoCo 3.3.2, LeRobot 0.6.1, seed 42), comparing PyTorch+AMP with ONNX Runtime CUDA Execution Provider (CUDA EP). The main evaluation uses 100 episodes/suite; a paired rollout uses 300 episodes/suite. PyTorch+AMP reaches 70.0%/88.0% Spatial/Object success at 1181 ms p99. Requested-FP16 and requested-INT8 ONNX reduce tether-inspect p99 to 601 ms and 532 ms, while Spatial success falls to 41.0% and 40.0% and Object remains at 89.0%. A graph audit shows those artifacts are byte-identical FP32 graphs, so the requested-INT8 row is not operator-level INT8 quantization. A static language-width ablation (16/24/32 tokens) yields Spatial success of 41.0%, 75.0%, and 71.0%; widths 24 and 32 recover much of the Spatial drop while Object success and uniform-bench latency stay approximately stable. Width-24 ONNX Spatial success is comparable to the PyTorch+AMP baseline at roughly half the latency (Wilson intervals overlap; two-proportion chi-squared p=0.53). Context width is an important contributor in this stack; it does not account for every PyTorch-vs-ONNX difference. Deployment evaluation should jointly report latency, artifact inspection, interface constraints, and closed-loop success. Code: https://github.com/rafiqul713/smolvla-libero-onnx.

Link Adaptation (LA) in 5G NR is inherently reactive, relying on channel measurements and HARQ feedback that may become quickly obsolete when the channel changes quickly. This data is also noisy, making it hard to track accurately, and has to be fed to real-time controllers with feedback-loop effects which are hard to troubleshoot. This explains why most practical deployments select simple but robust algorithms, which accept that the lag can leave the scheduler operating at overly aggressive or unnecessarily conservative rates, trading spectrum efficiency for predictable performance. In this paper, we improve on this status-quo with NOSTRAdAMUS, a predictive LA framework which adds foresight to existing algorithms without replacing or redesigning them. NOSTRAdAMUS predicts whether a retransmission will occur in the next radio frame from recent HARQ history, and applies corrections to the Modulation and Coding Scheme (MCS) selected by the underlying policy. We benchmark several ML models and show that Gradient Boosting achieves 82.9% accuracy overall with high-confidence interventions that are correct 94.2% of the time, and an inference latency of 5.5 {\mu}s. We train the model based on data collected Over-the-Air (OTA) on the X5G testbed, using the open-source OpenAirInterface (OAI) 5G stack, NVIDIA Aerial, and COTS O-RAN Radio Units and User Equipments. The model is then deployed as a dApp, which we evaluate OTA as well as on various channels with hardware-in-the-loop channel emulators. This includes 3GPP TDL and CDL channels, SISO and MIMO configurations, and pedestrian and vehicular mobility. Our evaluation shows that without retraining, and across this variety of scenarios, the dApp augments two SOTA LA algorithms, and increases goodput by up to 71.5% while reducing retransmissions by up to 71.8%. This demonstrates the robustness and generalization capabilities of our approach.

Streaming automatic speech recognition (ASR) for real-time voice agents and full-duplex dialogue must provide accurate partial transcripts with low commit latency. Existing systems commonly use a fixed chunk size, look-ahead, or target delay, or encourage emissions near estimated acoustic boundaries. These approaches do not directly optimize how much additional context to use at each output position under a single-pass, hard-commit constraint. We propose X2Streaming-ASR, which decomposes streaming recognition into when to commit and what to commit. Its three-stage training procedure first establishes streaming recognition ability, then warm-starts the commit policy with automatically probed trajectories, and finally refines the policy using character-level, segment-assigned group-relative rewards for recognition accuracy and latency. Across AISHELL-1/2/3 and WenetSpeech, X2Streaming-ASR achieves a mean character-level commit latency of 24-97 ms relative to forced-aligned character endpoints, compared with 409-585 ms for the evaluated streaming baselines. It achieves the best streaming CER among the evaluated systems on AISHELL-1 and AISHELL-3 with substantially lower latency.

Large language models (LLMs) are increasingly used as backends for intelligent web services, but serving them across the edge continuum requires balancing quality, latency, model footprint, and energy. This paper presents a controlled measurement study of self-hosted LLM inference across edge and near-edge deployment nodes: an NVIDIA Jetson AGX Orin and a near-edge server with CPU-only and GPU-enabled inference modes. We evaluate multiple open-weight LLMs and quantization variants using a fixed question-answering workload, and compare them against GPT-4o as a cloud-hosted accuracy and latency reference. Our benchmarking pipeline reports accuracy, model footprint, per-token decoding latency, prefill latency, and overall execution energy. The results show that GPU-enabled server execution provides the lowest compute-side latency, while Jetson Orin shows lower measured energy, consistent with its lower platform power under our setup. CPU-only execution is consistently dominated in latency for our workload and shows higher measured energy. We also show that parameter count and downloaded weight-file size alone do not reliably predict observed accuracy or latency. Finally, using Pareto-frontier analysis, we study how deployment decisions may change under possible streamed-token delivery overheads, highlighting that compute-side inference metrics alone can lead to suboptimal placement for latency-sensitive interactive web services.

This study investigates the classification of individuals as healthy or at risk of Parkinson's disease using machine learning (ML) models, focusing on the impact of dataset size and preprocessing techniques on model performance. Four datasets are created from an original dataset: DS_0, (normal dataset), DS_1 (DS_O subjected to Canny edge detection and Hessian filtering), DS_2 (augmented DS_0), and DS_3 (augmented DS_1). We evaluate a range of ML models-Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Gradient Boosting (GB), XGBoost (XBG), Naive Bayes (NB), Support Vector Machine (SVM), and AdaBoost (AdB)-on these datasets, analyzing prediction accuracy, model size, and prediction latency. The results show that while larger datasets lead to increased model memory footprints and prediction latencies, the Canny edge detection preprocessing supplemented by Hessian filtering (used in DS_1 and DS_3) degrades the performance of most models. In our experiment, we observe that Random Forest (RF) maintains a stable memory footprint of 61 KB across all datasets, while models like KNN and SVM show significant increases in memory usage, from 5.7-7 KB on DS_0 to 102-220 KB on DS_2, and similar increases in prediction time. Logistic Regression, Decision Tree, and Naive Bayes show stable memory footprints and fast prediction times across all datasets. XGBoost's prediction time increases from 180-200 ms on DS_0 to 700-3000 ms on DS_2 (truncated)

Enterprise wireless access points (APs) are promising platforms for predictive machine learning (ML), but their primary responsibility remains providing wireless connectivity and network services. Predictive inference must therefore share an AP's CPU and memory with packet processing, Wi-Fi and IoT radio operations, and client management. This resource contention creates two risks: a model that performs well on proxy hardware may be too slow on the target AP, while a model that fits in isolation may still degrade network services under load. We define \textit{network-aware deployability} using two gates: qualification of the model and its execution path on the target AP, followed by validation of its execution profile under packet-service and forecasting constraints. Our benchmarks show that edge testbeds do not reliably capture target behavior. Across matched artifacts and serving settings, five model implementations run 6.1--19.1$\times$ slower on an AP than on a Raspberry Pi5, while peak memory usage differs by up to 22%. Moreover, two forecasting foundation models of similar size differ in AP latency by 19$\times$. When serving a smaller model across 13 parallel streams at a 30s cadence under network saturation, default execution increases p99 round-trip time (RTT) by 76% and reduces throughput by 7.06%. Understanding these trade-offs is essential for live deployment if we aim to use APs for both networking and ML workloads.

Stream learning is commonly evaluated through predictive performance and adaptation to concept drift. However, sustained operation of a stream learner also requires predictable and bounded resource usage even on long streams. This requirement becomes even more critical when learning moves from servers to near-sensor embedded systems where memory and processing are scarce resources. In state-of-the-art stream learning, however, we perceive a strong focus on concept drift adaptation, whereas resource usage is often an evaluation byproduct. To close this gap, we benchmark seven representative stream classifiers on 13 real and synthetic streams under model-size budgets from 128,KiB to approximately 8,MiB. Our benchmark comprises a total of 6,463 experiments. We measure failure-aware accuracy, peak model size, time to budget exhaustion, and prediction-plus-update latency. The results reveal two distinct resource failure modes. Adaptive ensembles can exceed small budgets almost immediately because of their initial footprint, even when their size remains stable thereafter. Incremental trees can fit initially but grow throughout a long stream, with HoeffdingTrees (HT) and Extremely Fast Decision Trees (EFDT) increasing by median factors of 7.37 and 5.87. Explicitly compact methods remain the only viable option under the smallest budgets, but are usually overtaken as larger budgets make adaptive ensembles competitive. Hence, many state-of-the-art methods are only partially applicable in embedded systems or for long-running systems. We therefore call on the stream-learning community to make bounded resource usage a first-class design objective alongside drift adaptation, and propose concrete steps toward this goal, including an API through which stream learners can explicitly expose and respect resource budgets.

UAV on-board vision systems are widely used for different activities, including monitoring in no-fly zones. In this case, the vision-equipped UAV streams a video to a ground server where an operator assists its activities. The latency of video transmission has a profound impact on the effectiveness of the operator assistance. However, most techniques available for video transmission still incur significant latency costs. In this paper, we propose a graph convolutional neural network-assisted (GCN-Assisted A2C) deep reinforcement learning (DRL) system model to find the optimal pixel-correlated area of a suspicious object. We combine the Lagrangian dual form with gradient descent to prevent lack of convergence and over- and under-penalization constraint violation during latency optimization. The proposed system model sends a sub-group pixel-correlated area of the frame from the UAV to the server rather than the transmission of the whole video frame. The proposed framework utilizes the GCN model to explore hidden representations of feature-correlated groups of pixels. Moreover, the GCN supervises the A2C model, which selects a subgroup to enhance transmission latency, thus supervising the training of UAV actions in A2C. Experimental results show that GCN-assisted A2C reduces video frame transmission latency together with false detection rate in UAV vision systems over other DRL and state-of-the-art models.

World action models (WAMs) condition robot actions on predicted futures, but iterative video rollout increases deployment latency. We ask whether action generation requires the evolving rollout trajectory or only its future representation. Across four WAMs on all 40 LIBERO tasks, paired closed-loop interventions show that masking or reassigning future-cache values changes execution and reduces success, indicating sensitivity to future values and their assigned positions. For Joint and Cosmos-2, however, replaying one fixed final-clean key/value (K/V) cache nearly preserves unmodified execution, with $1.7$ to $1.9$cm end-effector average displacement error and $97.9\%$ to $98.2\%$ success. This separates cache consumption from production: these models can reuse a fixed cache but still require iterative rollout to construct it. We therefore propose RIFT (\emph{Rollout-free Imagination via Future Tokens}), which uses learned anticipation tokens to construct a complete future K/V cache in one backbone pass while retaining the original future-read interface. On LIBERO, RIFT achieves $98.8\%$ success, close to rollout-based Joint, IDM, and LingBot-VA at $98.4\%$ to $98.6\%$, while reducing action-chunk latency by $68.2\%$ to $89.1\%$. On RoboTwin2.0, RIFT reaches $92.9/92.6\%$ on clean/randomized scenes, the highest observed among the evaluated methods. These results support rollout-free future conditioning without iterative video generation at deployment.

Multilingual short-text classification supports operational systems such as content moderation, customer support routing, and intent recognition, yet aggregate evaluation often hides large differences between high-resource and low-resource languages. Uniform inference policies are simple to deploy, but they assume that all languages are equally well served. In this work, we evaluate a fixed-list routing strategy that keeps stronger languages on a direct multilingual path and selectively sends weaker languages through translation into English before zero-shot classification. The pipeline is fully self-hosted, uses pretrained compact sentence encoders, and requires no task-specific fine-tuning. We test the approach on two benchmarks chosen to differ in scale and label granularity: a 15-language subset of SIB-200 for seven-way topic classification and a 15-locale subset of MASSIVE for intent classification over an official 60-intent inventory. On SIB-200, the best overall configuration is R1, which translates only the low-resource tier: high-tier and mid-tier Macro-F1 remain unchanged, while low-tier Macro-F1 rises from 0.4632 to 0.6828. On the MASSIVE subset, the same low-tier intervention raises low-tier Macro-F1 from 0.2143 to 0.4417, but the best overall result is obtained by full translation, R3, at Macro-F1 0.4647. Across these two benchmarks, selective translation is a reliable intervention for weaker languages, whereas the optimal routing boundary depends on the task. We therefore report routing through tier-level quality gains and tier-level latency rather than a single global efficiency score.

Zero-shot text-to-speech (TTS) now supports interactive assistants, personalized media, and accessibility tools. All TTS systems require faithful linguistic rendering, consistent speaker identity, and low-latency response. Yet compact streaming systems must preserve sufficient acoustic detail in a predictable low-rate latent sequence, while iterative diffusion sampling and classifier-free guidance multiply inference cost at every autoregressive step. To strike a balance between high-fidelity synthesis and low-latency inference, we present CuteTTS, a compact continuous-autoregressive TTS system. It combines semantically aligned causal VAE latents with patch-level autoregression, explicit speaker conditioning, and a bidirectional flow-matching head. We further introduce guidance-step distillation, which absorbs classifier-free guidance and multiple solver steps into a single interval-conditioned student. Evaluations on LibriSpeech and Seed-TTS-Eval demonstrate competitive intelligibility and speaker similarity in zero-shot voice cloning, while distillation lowers first-audio latency by 23.3% and real-time factor by 40.8% relative to the base model with comparable objective and subjective quality. These results provide a practical path toward continuous-autoregressive TTS that reconciles high-fidelity generation with the latency demands of real-time interaction.

As LLM inference shifts to multi-tenant GPU clusters, co-batching improves throughput but obscures per-tenant usage and limits control. Enabling fractional sharing of the inference engine requires a real-time, per-request attribution primitive that is accurate and light enough to run inside the scheduling loop. We present LLMVisor, a roofline-guided latency attribution model that captures the memory-bound and compute-bound phases via a concise piecewise-linear form over features proportional to FLOPs and memory I/O traffic. LLMVisor decomposes batch latency into additive, per-request shares and runs efficiently at microsecond scale. We evaluate LLMVisor across Llama 3.1-8B and Qwen 2.5-14B/32B on A100/H100 GPUs under varying tensor parallelism and workload mixes. Compared to a token-count baseline, LLMVisor attains near-perfect R-squared and reduces relative error by up to 2.5x and 3.3x at p90 and p99, respectively, for prefill, and by up to 3.5x and 4.4x for decode, despite batching variability and sequence divergence.

Quantized Neural Networks~(QNN) with low-bitwidth data have proven promising in efficient storage and computation on edge devices. To mitigate accuracy degradation while maximizing speedup, layer-wise mixed-precision quantization~(MPQ) becomes a popular solution. However, existing algorithms for exploring MPQ schemes are limited in flexibility and efficiency. Comprehending the complex impacts of different MPQ schemes on post-training quantization and quantization-aware training results is a challenge for conventional methods. Furthermore, an end-to-end framework for the optimization and deployment of MPQ models is missing in existing work. To address these challenges, we propose the MiCo framework, a holistic MPQ exploration and deployment framework for edge AI applications. The framework adopts a novel optimization algorithm to search for accuracy-optimal quantization configurations under strict latency constraints. We further extended the framework to MiCoPro, which introduces a robust Hardware-Aware Proxy (HAP) model to enhance prediction accuracy and hardware versatility. By leveraging target-specific latency modeling, MiCoPro enables rapid exploration and direct deployment from PyTorch models to bare-metal C code. We demonstrate the versatility of our framework on both the BitFusion accelerator and SIMD-extended RISC-V processors, achieving up to 40% of latency reduction with less than 3% of accuracy drop.

The reasoning and agentic capabilities of large language models have expanded the range of applications they support, from short interactive exchanges to long, compute-heavy requests. LLM serving platforms today define response-latency service-level objectives, even though requests within the same service can differ by orders of magnitude in input length, generation length, execution cost, and the availability of reusable KV-cache state. As a result, requests governed by the same service level objective have different urgency: after accounting for the time required to execute them, some have substantial latency headroom while others have almost none. We define this headroom---the difference between a request's service level objective and its predicted remaining service time---as its per-request latency budget. We present Cascade, an LLM serving system that estimates and continuously updates this budget from request characteristics, KV-cache state, and current system load. Unlike prior SLO-aware schedulers that use deadlines to govern request ordering alone, Cascade uses a single per-request budget to jointly coordinate request scheduling and KV-cache management across the memory hierarchy. Its scheduler prioritizes requests with little remaining budget, while its memory manager uses the same budget to decide whether non-resident KV state should be restored or prefetched from a deeper tier, retained in HBM, or recomputed. By directing queueing and data-movement overhead toward requests that can absorb it, Cascade improves SLO-satisfied goodput while preserving fairness across heterogeneous request classes. On production traces across three large language models, Cascade improves goodput by up to2.4x and reduces SLO violations by 40% relative to the default vLLM first-come, first-served scheduler.

Time-Sensitive Networking (TSN) and Mobile Edge Computing (MEC) hold strong potential for enabling ultra-reliable low-latency communication for time-sensitive applications, such as eXtended Reality (XR). However, the widespread adoption of XR introduces significant challenges due to co-located services in MEC environments, leading to contention for shared network resources. Moreover, XR traffic types have distinct characteristics and criticality in terms of timing requirements, further increasing the complexity and dynamics of such environments. Although reinforcement learning has shown promise for TSN scheduling optimization in dynamic network scenarios, existing approaches rely on centralized or high-level multi-agent designs and are typically tailored to periodic and predictable industrial traffic, limiting their applicability to XR workloads. As a result, these approaches suffer from (i) limited ability to capture inter-queue dependencies due to coarse-grained control, and (ii) poor adaptability to highly dynamic and heterogeneous XR traffic. To address these gaps, we propose a multi-agent reinforcement learning approach for queue-level XR traffic scheduling. We adopt the multi-agent transformer (MAT) to model inter-queue dependencies via attention over agents' observations and actions, enabling implicit coordination across heterogeneous co-located XR applications. Our simulation results show that the proposed method outperforms baselines, achieving up to 71.42% latency reduction and up to 83.2% reduction in failure rate, while consistently achieving high reliability across all queues.

Warehouse items differ in how urgently they must be moved: perishable goods, pharmaceutical shipments, and just-in-time production materials must be delivered sooner than the rest of the stock. Decentralised robot swarms suit warehouses that cannot justify fixed automation infrastructure, but current swarm controllers treat all items alike or rely on an external scheduler to set priorities, so urgent items wait as long as ordinary ones. This paper presents a swarm logistics system in which each warehouse carrier holds an ultra-low-power Internet-of-Things (IoT) tag that broadcasts the urgency of its item over Bluetooth Low Energy (BLE). Robots read these broadcasts directly and weigh urgency against travel distance when choosing which carrier to serve, so prioritisation happens at the item level without central scheduling. The system is evaluated in simulation and validated on real robots and IoT-tagged carriers against a proximity-only baseline. In the physical trials, priority alignment (i.e. proportion of urgent items served first), improved from 0.41 to 0.64, with a nonsignificant trend toward lower 95th-percentile (P95) delivery latency and throughput within 1.2% of the baseline. In simulation, the benefit grew with system size: across three larger configurations, P95 latency fell by 5.2% to 11.8% and priority alignment improved by 41.7% to 51.6%. Attaching urgency to the items themselves therefore allows a decentralised swarm to serve time-critical stock sooner while keeping the low infrastructure requirements that make swarm systems attractive for warehouse automation.

As robotic systems become increasingly complex, generating control firmware from structural description files has emerged as a promising paradigm for reducing development complexity and improving maintainability. Existing robot description formats naturally represent robotic systems as hierarchical tree structures, where devices are recursively composed into functional subsystems and eventually into the complete robot. However, such tree-structured organization also introduces structured data dependencies that affect perception-to-decision latency and, consequently, control performance. In this paper, we propose FineMote, a control firmware generation framework with a scheduling mechanism tailored for tree-structured device models. The framework objectifies heterogeneous low-level control logic and exposes unified scheduling units and execution entry points. Based on the resulting object hierarchy, the scheduling mechanism exploits compile-time information to statically determine execution order with minimal runtime overhead. We prove that the proposed mechanism satisfies deadline and precedence constraints, and further derive an upper bound on intra-tree decision latency. We implement the proposed framework and evaluate it on real robotic control platforms. The experimental results show improved timing behavior and runtime responsiveness, demonstrating the practical effectiveness of the proposed design.

Agentic AI is emerging in datacenters, but its architectural implications remain unexplored. We organize agentic workflows in a taxonomy and present its first architectural characterization with a production study at Microsoft Azure and a controlled study of open-source frameworks. We show that agentic execution is fragmented and heterogeneous. Requests expand into a workflow of LLM inferences, tool invocations, and orchestration decisions that repeatedly cross the CPU-GPU boundary. Our taxonomy explains how this fragmentation turns into resource demand. As orchestration and tools run on the host, the CPU sits on the critical path. Execution structure sets the load over time, which stays low with sudden spikes. Model composition sets how evenly the workflow uses the GPUs. Diversity in tasks and tools widens this range even further. These characteristics expose architectural mismatches of conventional uniform servers. Fragmented execution strands CPU and GPU capacity despite bursty demand. Different software roles make homogeneous CPU provisioning inefficient. Finally, multiplexing many agents onto shared cores degrades microarchitectural locality. Guided by our findings, we derive implications for agentic servers and examine them through Agora, our prototype for commodity servers. Agora dynamically harvests idle CPU cores for co-located throughput work, while protecting agentic tail latency against tool spikes. It oversubscribes GPU memory by placing more agents on each GPU, prefetching the next agent's state to hide swap latency. To match the machine to the heterogeneous roles, Agora pools cores by role and applies affinity-aware scheduling to restore locality. It automatically tunes mechanisms to the workload. Agora improves utilization and server throughput while preserving agent tail latency. Our insights also identify key directions for future server architectures for agentic AI.

Onboard satellite intelligence requires a task layer that translates mission intent into local tool calls, exposes execution state, and returns machine-consumable artifacts under communication and power constraints. We present SAT-Edge-Agent, a hardware-in-the-loop (HIL) edge-agent system deployed on a commercial off-the-shelf ARM-based heterogeneous edge system-on-chip. A browser workspace and FastAPI agent coordinate a local OpenAI-compatible language service with a project-internal YOLO-style oriented-object-detection endpoint that returns FAIR1M metadata-backed structured results. Two fixed FAIR1M workloads, one single-image and one serial two-image request, were repeated 20 times each and completed 20/20 attempts. Mean Full-Agent latency was 29.353 s and 60.937 s, with empirical P95 values of 31.166 s and 66.882 s. Mean detector time was 861.386 ms and 1510.920 ms, only 2.93% and 2.48% of the corresponding Full-Agent means. Profiling indicates that most visible latency occurs outside detector execution. Mean CPU utilization was 20.761% and 20.482%. A 200-ms NPU-load field averaged 100% for both workloads, but it represents a shared-accelerator software field rather than detector-only occupancy or calibrated utilization. The public evidence package provides sanitized request-level records, redacted JSON, normalized SSE examples, and scripts reproducing the reported statistics. These results establish a reproducible HIL boundary for observable satellite edge-agent orchestration, but do not establish detector accuracy, a new geolocation method, calibrated energy efficiency, or flight readiness.

Fewer visual tokens do not guarantee lower end-to-end latency. We evaluate break-even with a reproducible protocol that accounts for decision overhead, shared work, and the operators each policy can avoid. A stage-level decomposition reconciles these components with measured end-to-end latency. In a 30-example pilot, the two tested autoregressive probes remain slower than Full despite state reuse. A lightweight post-vision predictor yields paired confidence intervals below zero on RTX 3090 and A100 and remains significant after a conservative all-pairs Holm correction. A pre-vision image-size rule also yields intervals below zero on both GPUs, although neither comparison remains significant after the same correction. Pre-vision routing has a structural opportunity unavailable to post-vision pruning: it can avoid preprocessing and vision encoding. On A100, this opportunity outweighs a nearly eightfold larger downstream token reduction by the post-vision policy. Reported quality is conditional on examples answered correctly by Full and is not benchmark accuracy.

Fast dynamic obstacle avoidance (DOA) on uncrewed aerial vehicles (UAVs) demands not only low-latency control and actuation but also reliable perception with sufficient sensing range for accurate obstacle detection and speed estimation. This letter presents, to the best of our knowledge, the first mmWave RADAR-based perception-and-control system for fast onboard DOA. We derive and analyze latency and spatial bounds that relate sensing range, relative speed, and control delay, yielding sufficient conditions for successful avoidance. Our system adopts a lightweight tracker based on interacting multiple models and a controller based on control-barrier functions that directly outputs evasive accelerations. It achieves position errors of less than 0.15 m, 0.93 m, and 0.87 m in x, y, and z directions for 300 experiments with three different object sizes and varying visibility (light and dark), and a similar spread for 90 experiments in smoke. An onboard implementation on a Raspberry Pi 4B demonstrates real-time feasibility with an end-to-end sensing-to-command latency of approximately 14 ms. Code and the full dataset of 390 throws are available (https://tinyurl.com/radardoagit).

Can large language models generate not just correct, but fast hardware? This paper investigates the question in financial FPGA design, where 5-10 nanoseconds of latency determines competitive advantage and designs iterate continuously as protocols, strategies, and regulations evolve. FinHardBench, a benchmark of 33 financial computing tasks, is presented together with three experiments that mirror the real-world FPGA iteration cycle: generating new modules from specifications, tuning system-level configurations across a 6-stage trading pipeline, and adapting existing modules to specification changes. Evaluation of six LLMs on 1530+ experiment rounds yields three findings: (1) models achieve 19-61% functional correctness with timing degradation up to 13.7$\times$ on specific tasks; (2) in system-level design space exploration, top LLMs converge to the optimal configuration with higher reliability than random search, simulated annealing, and Bayesian optimization baselines (5/5 seeds vs. 0-4/5 at the same 24-round budget); (3) strategy-level specification changes remain unsolved for most models. Across the six models, generation and DSE rankings overlap moderately: the strongest code generator is not the fastest architecture optimizer, and the weakest code generator (MiniMax M2.7) still reaches the system optimum on 4 of 5 seeds. On the tasks in FinHardBench, difficulty tracks training data pattern availability more closely than abstraction level. FinHardBench is released as an open-source benchmark.

Schema linking is a critical component of Text-to-SQL systems, but existing approaches often trade off contextual modeling capacity, score-based controllability, and inference efficiency. We introduce AttnLink, an attention-based framework that converts LLMs' internal attention into continuous relevance scores for schema items. AttnLink extracts the attention from the generation-start position to candidate schema spans, enabling all candidates to be ranked in a single prefill pass without autoregressive decoding. We develop two variants: AttnLink-U, which directly probes pretrained attention without parameter updates, and AttnLink-S, which aligns the attention distribution with gold schema items through direct supervision. To improve coverage of multiple relevant schema items, AttnLink-S combines a set-mass objective with an adaptive probability-floor regularizer. The resulting scores support post-hoc precision-recall control through temperature scaling and cumulative-mass selection. Experiments on Spider, BIRD, and Spider2-SQLite show that AttnLink-S achieves mAP scores of 99.22%, 95.95%, and 83.29%, respectively, with millisecond-scale schema-linking latency. It also yields the best or tied-best execution accuracy for downstream SQL generation in seven of nine generator-dataset settings.

We present a real-time upper-body human-to-humanoid motion imitation framework driven by neuromorphic event-based vision. This work addresses practical perceptual bottlenecks of conventional frame-based RGB sensors, specifically their difficulty in high dynamic range (HDR) scenes and rapid motions due to fixed integration times. By leveraging the Prophesee EVK4 event camera, which operates asynchronously with high temporal resolution and a dynamic range exceeding 120 dB, our system supports stable tracking in conditions where standard vision pipelines degrade, such as severe backlighting and very low light environments below 5 lux. The architecture integrates a low-latency Perception Module, utilizing optimized event accumulation and gravity-aligned inertial fusion, with a causal Motion Module (TWIST) that performs online kinematic retargeting. We validate the system on an embedded NVIDIA Booster T1 platform and an 18-DoF humanoid upper-body setup, demonstrating an end-to-end photon-to-action latency of 23-34 ms and advantages over RGB baselines under our experimental setup. The results indicate a practical trade-off: events can be preferable for fast or poorly lit upper-body teleoperation, whereas well-lit static scenes may favor RGB or hybrid sensing.

AI coding agents like GitHub Copilot, Claude Code, and Codex interleave multi-step LLM inference with tool execution, creating a workload different from chatbots. We present the first production-scale characterization of this workload using sampled GitHub Copilot traces from June 2026, comprising 3.2M users, 13M sessions, 761M LLM calls, and 95T tokens. Our analysis reveals distinctive workload properties with important systems implications. For example, agentic coding sessions consist of sparse user-initiated turns, each unfolding into an autonomous agent loop of LLM calls almost always coupled with tool execution. This structure yields KV cache hit rates averaging 90% within a turn, but falling to 55% across turn boundaries and drastically invalidated after events like model switches or context compaction. Diverse workflows and user behaviors are observed with variable and long-tailed token consumption, time span, and tool calls. We highlight the difference between quick agentic turnaround times and the minutes-long user idle periods at turn boundaries, and design a lightweight idle-time predictor that captures 86-90% of total idle time, enabling proactive decisions for efficient resource orchestration. These findings challenge assumptions underlying current LLM-serving systems and provide an empirical foundation for agent-native infrastructure.

Long-form and real-time talking-head generation remains challenging due to a latency-quality trade-off: inefficient multi-step diffusion prohibits streaming generation, whereas real-time autoregressive approaches suffer from error accumulation and identity drift. To address this drawback, we propose LeapTalk, a novel framework that achieves stable and real-time talking-head generation with a single forward step, scaling to arbitrarily long videos. At the heart of our approach lies a single-step bridge distillation scheme. On the one hand, departing from the conventional noise-to-data paradigm, we introduce a data-to-data transport formulation based on a Brownian bridge. Anchored by a persistent reference, this strategy effectively mitigates identity drift and enhances long-term temporal stability. On the other hand, to enable smooth knowledge transfer from a pre-trained diffusion teacher to the student bridge model, we explore a heterogeneous distillation framework with an SNR-aligned time transformation $Φ(τ)$, which bridges the functional discrepancy between the two models. Moreover, we propose an audio-driven classifier-free guidance mechanism to maintain fine-grained lip synchronization under extreme step reduction. Extensive experiments demonstrate that our method achieves high-fidelity and temporally consistent video generation with only 1 step at up to 200 FPS, significantly outperforming existing approaches in both efficiency and stability. Project Page: https://zhangrongxiang.github.io/leaptalk-page/

Deploying learned control policies on low-cost robotic platforms introduces transport latencies and noisy motor feedback that systematically widens the sim-to-real gap. The chasm of simulation to deployment in hardware lies in the delay of the actuator reaching the commanded position. On platforms such as the Mini Pupper 2, a measured >50 ms transport delay transforms the locomotion task from a standard Markov decision process into a partially observable one. In this paper, we take a biologically inspired approach of handling noisy and delayed feedback to close the sim-to-real gap, thereby expanding the capability of reinforcement learning on cost-constrained hardware. Using a low-cost quadrupedal hardware platform, we find that using a forward model of the average actuator delay, paired with a time-aware neural network results in robust locomotion. Additionally, our time-aware neural network learned a central pattern generator (CPG): a self-sustaining rhythmic gait that is robust to +320 ms latency perturbations, mirroring the CPGs found in the spinal cords of vertebrates. We posit that temporal self-organization may be a general strategy for cost-constrained locomotion.

Human-Robot Collaboration (HRC) in shared physical spaces using Augmented Reality (AR) interfaces is powered by Simultaneous Localization and Mapping (SLAM). Existing multi-agent SLAM systems rely on an edge server to combine visual findings of multiple resource-constrained agents, perform computation, and schedule updates to their local maps. However, the edge treats all agents uniformly and ignores the fundamentally different latency requirements of heterogeneous HRC agents: robots and head-mounted AR users. This uniform resource allocation often results in high lag for user manipulation, as it does not meet the stringent latency requirements of AR. In this work, we design, implement, and evaluate SHARE, a user-centric SLAM system that strategically prioritizes AR user experience while maintaining accurate tracking performance for robots. SHARE builds a first-of-its-kind experience model for HRC agents and adaptively adjusts transmission priorities to match it. To reduce end-to-end latency, SHARE leverages the redundancy of visual features acquired by agents in shared human-robot workspaces to reduce computation time induced by edge-based processing. Real-world deployment with commercial AR headsets and a ground robot achieves 13.22 ms average latency for AR users (43.3% reduction from baseline) while maintaining sub-2-centimeter tracking accuracy. User studies further reveal statistically significant improvements in user perception.

Large language model (LLM)-based dialogue systems suffer response delays because generation begins only after final speech recognition. While fixed fillers are a workaround, they become unnatural over time. We propose a two-stage incremental framework that decouples prefatory-response preparation from speech onset. Once user intent becomes predictable, an intent readiness detector triggers LLM-based generation of a short prefatory response. Concurrently, a voice activity projection (VAP) model determines when to deliver it. Through a field experiment with a route-guidance robot in a shopping mall, we evaluated three conditions: no-filler, fixed-filler, and contextual-preface. Both fixed-filler and contextual-preface significantly reduced initial response latency relative to no-filler. Relative to fixed-filler, contextual-preface had significantly longer initial response latency but a significantly shorter initial-to-main gap. Exploratory ratings showed no significant differences. These results indicate a timing trade-off.

AI agents are enabling a new paradigm of agent-augmented real-time communication (RTC), where humans focus on high-level collaboration, while agents autonomously retrieve, analyze, and generate information in real time to support their interactions. These apps enable new experiences across various domains: for example, when corporate employees co-author a legal document, their agents can discuss and draft on their behalf, sparing them the burden of manually reviewing each other's work. As existing cloud-based agents suffer from privacy risks and unscalable server costs, on-device agent-augmented RTC offers a promising alternative. However, this on-device paradigm introduces a new networking challenge: contention between concurrent traffic flows generated by humans (for live video streaming) and agents (for sending context files for analysis). We design HFS, a framework to ensure both high live video quality and low agent response latency in agent-augmented RTC apps. We achieve the goal through an app-guided multi-flow transport approach, where a unified app-layer orchestrator jointly controls the sending rates of live video and agent context flows based on their heterogeneous app requirements. Our prototype built atop WebRTC and llama.cpp demonstrates that HAFS outperforms baselines, achieving 1.5x higher video quality while reducing agent response time by 31%.

Routing to select large language models (LLMs) with different cost-quality trade-offs has become a fundamental deployment feature of enterprise AI. Existing routers, primarily make independent routing decisions for each LLM call. However, agentic applications execute as long-horizon workflows whose quality is determined only by a delayed, task-level outcome. This mismatch prevents per-call routers from correctly attributing feedback to individual routing decisions. Towards mitigating this, we present TRACE-Router, a task-level routing framework that aligns routing with the unit of supervision. TRACE-Router assigns each task to a model once at admission using a contextual bandit, pins all subsequent LLM calls to the selected backend, and updates its policy using the task's terminal reward, jointly accounting for accuracy and latency. By leveraging delayed task feedback, TRACE-Router learns routing policies that adapt to the workload while avoiding explicit task-complexity estimation. Across three agentic benchmarks, TRACE-Router consistently improves the accuracy-latency trade-off, achieving non-dominated Pareto frontier points. On tau2-Bench, it outperforms latency-matched interpolation between individual models by 7-8 accuracy points, while on Terminal-Bench it achieves 7.1 higher accuracy points than the strongest single model baseline with 36% lower latency.

Interactive deep image segmentation enables efficient medical image annotation by iteratively refining predictions from user prompts, such as positive and negative clicks. Recent patch-based methods, including nnInteractive, achieve strong segmentation performance but remain limited in annotation workflows by high interaction latency, limited responsiveness to successive interactions, and the lack of support for reversible prompting. Furthermore, evaluation relies predominantly on simulated rather than controlled real-user interaction studies. We present SLIP, an end-to-end trainable framework for interactive 3D medical image segmentation that decouples image encoding from prompt-guided refinement. Image features are computed once and reused, while a lightweight patch memory bank maintains an interaction-aware segmentation state shared across patches. This representation enables prediction updates by propagating interaction context throughout the image, supports reversible prompting without recomputing image features, and substantially reduces interaction latency. By separating image representation from interactive reasoning, SLIP remains compatible with a wide range of image encoders. We train a single SLIP model for general interactive segmentation across diverse anatomical structures and imaging modalities. Beyond standard simulated evaluation, we conduct a controlled prospective user study comparing manual segmentation, nnInteractive, and SLIP across three clinical annotation tasks, six expert participants, and subjective usability measures, addressing the limited human validation of interactive segmentation methods. SLIP achieves SOTA interactive segmentation performance across 13 public datasets while providing lower interaction latency, greater responsiveness, support for reversible prompting, and higher user preference than existing approaches.

A two-stage mixed-integer linear programming framework is introduced for subsea pipeline incident response planning, jointly optimizing Subsea Docking Plate (SDP) placement and resident autonomous underwater vehicle allocation to minimize both maximum and average response times under spatial uncertainty. Phase 1 minimizes the maximum response time across all potential leak locations. Phase 2 reduces the average response time subject to the maximum bound. A case study based on the Johan Sverdrup oil and gas field in Norway, with 9 SDP options and 33 pipelines, shows that just a few well-placed SDPs are enough to achieve top performance. A cost versus time Pareto analysis reveals the trade-off between deployment expenditure and response efficiency, providing actionable guidance for designing resilient and cost-effective subsea inspection networks.

The trade-off between robustness, latency, and prosody critically challenges text-to-speech (TTS) systems. Autoregressive models, despite fidelity, are slow and error-prone; non-autoregressive (NAR) alternatives, while fast, often sacrifice prosodic naturalness via rigid alignments. This paper introduces StellarTTS, a novel mobile-optimized NAR TTS framework based on a sparse temporal embedding strategy, enabling granular control of phoneme duration, pronunciation, and prosody. Furthermore, we propose a semantic-aware codec that facilitates efficient single-stage decoding. Conditioned on the sparse temporal embedding, our 83M-parameter lightweight masked generative transformer achieves a real-time factor (RTF) of 0.08. Experiments demonstrate that StellarTTS attains lower latency and stronger robustness compared to state-of-the-art TTS systems, while maintaining competitive performance in audio quality, prosodic naturalness, and speaker similarity.

Federated learning (FL) over wireless networks suffers from significant training latency and degraded convergence due to unreliable wireless transmission, especially under blocked propagation environments. Although reconfigurable intelligent surfaces (RISs) can improve communication reliability, existing wireless FL studies rarely characterize the trade-off between learning convergence and communication delay under modulation-dependent transmission errors. In this paper, we consider a wireless FL system operating under RIS-assisted blocked-link propagation scenarios, and focus on adaptive modulation and sub-channel allocation for convergence-latency aware communication design. By characterizing the effect of symbol errors on uploaded local gradients, we derive a convergence-related upper bound that reveals the impact of symbol error rate (SER) on FL loss decay. Based on this result, we formulate a joint convergence-latency optimization problem, which is cast as a mixed-integer nonlinear programming (MINLP) problem, and solve it using a low-complexity hybrid alternating optimization framework. Extensive experiments on MNIST, CIFAR-10, and Speech Commands show that the proposed scheme consistently achieves faster convergence and higher test accuracy than existing adaptive communication schemes, especially in complex tasks and challenging wireless scenarios.

Autonomous racing exposes the sim-to-real gap under extreme operating conditions characterized by high speed, tight stability margins, and stringent real-time constraints. Although simulation is indispensable for development, controllers that perform well in simulation often degrade abruptly on physical platforms due to interacting effects of dynamics mismatch, estimation delay, and execution-layer latency. This paper frames sim-to-real transfer in autonomous racing as a full-stack, real-time systems problem. We introduce a structured three-layer perspective (Physical/Cyber/Execution) to analyze how mismatches propagate and amplify through closed-loop feedback. We present diagnostic metrics beyond nominal lap time, including performance flip, stability-oriented measures, sensitivity to delay and noise, and latency distribution characterization. Mitigation strategies are synthesized from a deployment-oriented viewpoint, emphasizing execution-aware and delay-aware design. Finally, we outline benchmarking guidelines that enable reproducible and fair sim-to-real evaluation under compute and timing constraints. The resulting framework clarifies cross-layer failure mechanisms and provides practical design principles for deployable autonomous racing systems operating near dynamic limits.

Single-stream autoregressive decoding of large language models is bound by memory bandwidth: each generated token requires one full forward pass through the target model, and successive passes cannot be parallelized. Speculative decoding restructures this computation: a small draft model proposes $K$ tokens autoregressively, the target model scores all of them in one batched pass, and a rejection-sampling rule provably preserves the target model's output distribution. We present a from-scratch, device-agnostic (CUDA/MPS/CPU) implementation and an empirical study across five draft/target backend configurations on a consumer Apple-silicon laptop. Distribution equivalence is verified at three levels, culminating in a two-sample test over roughly 9,200 real-model tokens per method ($χ^2 = 162.5$, dof $= 200$, $p = 0.976$) and exact greedy-sequence agreement. The best configuration reaches a measured $1.61\times$ wall-clock speedup at $K=6$, on an acceptance profile declining from 69.7% at $K=1$ to 37.8% at the optimum, while three of five configurations decelerate, either because the draft fails to out-speed a small target or because the quantized Metal backend executes "parallel" verification serially, an effect we isolate and quantify. The failures are as instructive as the successes: speculative decoding pays off only when verification is genuinely batch-parallel and the draft/target latency gap is real.

Stream-processing systems increasingly operate across heterogeneous mobile edge--cloud infrastructures, where workload volatility, resource contention, and stringent quality-of-service (QoS) requirements complicate decentralized scheduling. This paper proposes \emph{MAS-DecStream}, whose main contribution is \emph{LLM-MR-CNP}: an extension of the classical Contract Net Protocol with semantic CFP formulation, progressive context disclosure, multi-round proposal revision, negotiation memory, and deterministic validation. Edge-cluster agents refine natural-language offloading proposals from local observations, predicted resource states, and qualitative runtime context, while hard resource and QoS constraints remain deterministic. Experiments derived from the Alibaba ASI Trace evaluate the extension at three levels: single- versus multi-round CNP, rule-based versus LLM-assisted refinement, and fixed-model single- versus multi-round negotiation. Under the evaluated configurations, MAS-DecStream reduces latency violations to 3%, eliminates resource overcommitment, reaches a conflict-resolution rate of 0.91 with 20 agents, and improves utility by up to 22% over the multi-round rule-based baseline. A separate 25-case evaluation shows model- and prompt-dependent accuracy--cost trade-offs. The results provide initial evidence that multi-round CNP refinement is the principal protocol-level gain, with LLM assistance adding value for qualitative and uncertain runtime context.

Throughput prediction is foundational for artificial intelligence-driven 6G resource orchestration. Conventional monolithic machine learning models struggle to generalize across diverse operators, mobility modes, and traffic types, leaving a critical stochasticity gap between signal conditions and achievable throughput. To overcome these constraints in heterogeneous urban environments, we propose a Tiered Multi-Agent System (TMAS) that dynamically routes edge telemetry to context-aware Domain Micro-Agents, validated on a dataset of 48,618 samples collected in Sunway City, Malaysia, with Nemo Handy drive test software, spanning three Tier-1 mobile network operators, three mobility modes, namely (i) elevated pedestrian walkway, (ii) ground-level shuttle bus, and (iii) elevated bus rapid transit; and three traffic profiles, namely (i) persistent download, (ii) persistent upload, and (iii) adaptive video streaming. Our evaluations reveal that TMAS overcomes predictability bottlenecks, achieving a coefficient of determination (R2) of up to 0.931 and a Mean Absolute Error (MAE) as low as 0.53 Mbps. The system demonstrates high operational efficiency, with rapid micro-agent training times, low inference latencies, and agentic routing overhead of 0.004 to 0.126 ms. These latency characteristics indicate the architecture is a promising candidate for the response times required by next-generation wireless networks.

In production large language model (LLM) deployments, high API availability guarantees do not equate to conversational continuity. When a primary provider experiences an outage or strict rate-limiting, naive stateless failover mechanisms successfully maintain uptime but silently discard conversation history, severely disrupting the user experience. To rigorously quantify and resolve this failure mode, we introduce two novel metrics: Continuity Preservation Rate (CPR) and Continuity Latency Overhead (CLO). We propose a stateful, multi-provider proxy architecture utilizing a History-Forwarding strategy to seamlessly reconstruct conversational state across heterogeneous LLM endpoints during failover events. Furthermore, we release continuity-bench, https://github.com/Vishal-sys-code/continuity-bench, an open evaluation harness designed to stress-test context preservation under high-concurrency provider failure conditions. Our empirical evaluation ($N=750$ failover events) demonstrates that our stateful proxy achieves a 99.20% CPR [95% CI: 98.27%, 99.63%], cleanly transferring deep conversational context to fallback providers, compared to a near-0% preservation rate for standard stateless architectures. Finally, we characterize failover latency distributions, identifying the critical necessity of asynchronous exponential backoff with jitter to prevent cascading retry storms against strict-limit fallback APIs. Our results provide a principled foundation for building robust, state-preserving multi-model inference systems.

We present Wan-Streamer v0.3, which reframes our native-streaming interaction model under a single organizing view: a video is a world plus an event stream. The world is the persistent context in which a video unfolds, including the environment, scene, subjects, ambient acoustic conditions, voice characteristics, and other relatively stable conditions. The event stream is everything that changes over time within that world, including scene or environmental changes, subject behavior, speech, and other sounds. This yields a general-purpose pretraining task over large amounts of real video: given a world and incoming input, predict how the world moves, changes, and responds in real time. The resulting competence can be specialized to a broad family of real-time downstream tasks. We instantiate it on real-time full-duplex audio-visual interaction, where the event stream is the agent's speech together with free-form behavior. Functionally, the model's multimodal understanding process is vision-language-action-like: it maps multimodal user input to language-form speech and behavior actions. Wan-Streamer v0.3 preserves the v0.2 operating point: 640x368 video at 25 FPS, a 160 ms streaming unit, approximately 200 ms model-side response latency, and approximately 550 ms total interaction latency under a 350 ms bidirectional network budget.

Reliable, low-latency uplink connectivity is a key requirement for C-V2X networks in dense urban environments, where fast channel variations and blockages often degrade direct vehicle-to-infrastructure links. Multi-hop relaying can restore coverage, but relay-link activation under radio, capacity, and routing constraints results in an NP-hard optimisation problem, typically solved via Mixed-Integer Linear Programming (MILP), whose runtime scales poorly with graph size. This paper introduces an edge-aware Learning-to-Optimise framework for real-time relay selection. Each V2X snapshot is modelled as a directed graph: node features encode vehicle state and traffic demand, while edge features capture radio-link capacity. An offline MILP oracle generates optimal relay configurations that supervise a Graph Isomorphism Network with Edge Features (GINE), enabling edge-level relay activation through a single forward pass, with tightly bounded inference latency. To bridge learning and exact optimisation, we also propose a hybrid GINE-Pruned MILP (GP-MILP) strategy in which GINE predictions prune the MILP search space. Experiments on a large-scale dataset generated via an OSM-SUMO-GEMV$^2$ pipeline show that GINE closely matches MILP decisions at the link level (accuracy 0.9589), F1-score (0.9544) on validation) and yields consistent end-to-end connectivity gains over a 1-hop MILP baseline (up to 9.2% with four RSUs and 12% with two RSUs). Inference latency remains tightly bounded, with all evaluated instances completing within 5ms. Moreover, GP-MILP preserves MILP-equivalent solutions (same objective value) while achieving solver runtimes below 30ms for more than 98%) of the graph instances, making MILP-grade optimisation compatible with stringent NR-V2X latency budgets.

Diffusion large language models (dLLMs) offer a theoretical advantage in parallel generation over standard autoregressive models. However, parallel generation alone does not guarantee practical speedups. Realizing this efficiency requires specialized inference mechanisms, such as diffusion-aware caching and reuse. Consequently, as inference efficiency becomes a prerequisite for practical deployment, recent research has actively explored acceleration techniques across algorithms, architectures, and systems. However, rigorous comparisons remain difficult, as end-to-end latency stems from intricate trade-offs between algorithmic, architectural, and system-level factors that are often conflated in existing benchmarks. In this survey, we introduce a unified latency decomposition framework for dLLMs to disentangle these factors and analyze their impact on inference speed in real deployments. Guided by this framework, we categorize acceleration techniques along three axes covering algorithmic innovations, architectural and system optimizations, and inference-time scaling. Finally, we provide guidelines for reproducible benchmarking and highlight open challenges for realizing the full potential of parallel generation.

With deep neural networks (DNNs) increasingly deployed on edge devices, hardware (HW)-aware optimization techniques--such as HW-aware compression and HW-aware neural architecture search (HW-NAS)--have become essential. These methods rely on real feedback from the target hardware to tailor DNN architectures for efficient deployment. While the search can be parallelized, latency measurements via hardware-in-the-loop (HIL) remain a bottleneck due to their sequential nature. Recent approaches use latency predictors to replace costly HIL feedback, but challenges persist: (1) platform-specific predictors often require tens of thousands of samples, and (2) inaccurate predictions can mislead the NAS process. To address this, we introduce HiFi-LLP, a high-fidelity, low-cost latency predictor based on graph attention networks, augmented with a confidence metric. HiFi-LLP outperforms prior platform-specific predictors by up to 9 percentage points (p.p.) in the 10% accuracy bound and achieves a Spearman's rank correlation of up to 0.996 across six devices in the LatBench dataset. We further propose a hybrid NAS framework that routes low-confidence predictions to HIL, achieving up to 8.6$\times$ speedup compared to typical NAS while maintaining a competitive Pareto front.

LLMs are now proposed for fraud detection, scam investigation, content moderation, and other trust-and-safety workflows. Much of the public literature still evaluates them as models, with less attention to their behavior as components in operational pipelines. This creates a practical evidence question: what would justify placing an LLM inside a live workflow with latency, cost, escalation, human-review, and adversarial-risk constraints? We address this question through a fraud-first survey of deployment evidence. We code 49 operationally relevant sources on LLM use in fraud detection, investigation support, content moderation, and cross-cutting robustness (18 fraud, 14 moderation, 17 cross-cutting), supplemented by 15 contextual references that establish the survey boundaries. These sources include systems, benchmarks, frameworks, and deployment-relevant surveys, not 49 production deployments. The main finding is an evidence imbalance. Fraud supplies the largest task-specific portion of the coded corpus. The moderation papers, however, include more explicit public evidence on latency, cost, governance, and fairness. Among the 18 fraud and investigation sources, none report clean per-decision latency, per-decision dollar cost, or calibration evidence; most report offline task performance, retrieval gains, or case-study accuracy instead. The survey contributes a role-and-evidence organizing frame, FORTE, for locating LLMs as classifiers, retrieval interfaces, explanation generators, reviewer assistants, agents, feature extractors, or escalation components. It also contributes a minimum deployment-evidence checklist covering latency budget, cost per decision, decision threshold, explanation integrity, and adversarial pressure. The resulting agenda identifies studies needed to support deployment claims for LLM-based fraud and trust-and-safety work.