Real-Time Inference

Embedded global navigation satellite system (GNSS) interference monitoring requires fast and memory-efficient inference to process large volumes of raw in-phase and quadrature (IQ) samples in real time. At the same time, increasingly expressive deep neural networks (DNNs) are needed for robust interference classification and characterization across diverse signal conditions. This creates a fundamental tension between predictive performance and deployability on resource-constrained hardware. In this paper, we investigate efficient network inference for GNSS interference characterization using iterative structured pruning, post-training static quantization, and hardware-aware zero-shot neural architecture search (NAS). Starting from MCUNet as a compact baseline, we analyze how model compression and automated architecture optimization affect model size, computational complexity, and memory usage while maintaining task performance. Experiments on a GNSS interference dataset, covering both classification and generalized characterization, show the benefits of combining compression and hardware-aware design for embedded deployment. Our results provide practical guidance for developing compact machine learning (ML) models for real-time GNSS interference monitoring on embedded platforms (iMXRT1062 MCU, Raspberry Pi Zero 2W, and Raspberry Pi 5).

Remote photoplethysmography (rPPG) estimates the blood volume pulse (BVP) signal from facial videos, enabling contact-free health monitoring. Conventional clip-wise approaches, which use video clips as input, require capturing over one hundred frames before inference, thus introducing several seconds of delay and hindering real-time use. Meanwhile, frame-wise approaches struggle to capture long-range temporal and periodic features of physiological rhythms, and therefore lead to reduced estimation accuracy. To overcome these issues, we propose StreamPPG, a unified architecture that enables low-latency frame-wise physiological signal estimation while achieving competitive accuracy compared with clip-wise approaches. StreamPPG is trained under a consistent privileged learning (CPL) strategy, which leverages ground-truth rPPG signals as privileged information to enhance the model's representation capability. Extensive experiments demonstrate that StreamPPG achieves state-of-the-art accuracy across multiple datasets while maintaining real-time throughput on edge devices.

Action chunking enables robot policies to produce temporally coherent behavior, but generating multi-step action sequences with flow-based policies incurs latency that is incompatible with real-time control. Under asynchronous execution, the robot continues executing the current chunk while the next one is generated, causing even minor delays to create inconsistencies at chunk boundaries. Existing methods address this problem by steering generation toward the already executed action prefix. We instead show that prefix consistency can be achieved by selecting an appropriate initial noise before generation begins, allowing the unmodified flow ODE to produce a coherent next chunk. This reframes asynchronous inference as a noise selection problem rather than a trajectory steering problem. We introduce \textbf{PAINT}, a training-free method that finds this noise via backward Euler inversion and constructs the final chunk through a repainting rule. In summary, \texttt{PAINT} requires no gradients, retraining, or policy modification; yet it improves execution consistency and task performance across \textit{12 simulated benchmarks} and \textit{6 real-world manipulation tasks} spanning single-arm, bimanual, and humanoid embodiments. Website: ~\href{https://paint-action-chunking.github.io}{\texttt{https://paint-action-chunking.github.io}}.

Deploying vision transformers (ViTs) on sensor-edge systems is limited not only by on-device compute, but also by the energy and bandwidth required to transmit high-dimensional image data from the sensor to the processor. While in-sensor and near-sensor computing reduce this cost through early feature extraction, existing methods often provide only modest compression. We observe that the frequency domain provides a naturally compact representation of visual information and can be exploited at the sensor level to reduce sensor-to-processor data movement. Building on this insight, we present FrequencyFormer, a co-designed sensor-to-processor pipeline for efficient ViT inference. FrequencyFormer includes: (1) a multi-scale DCT tokenizer that compresses a 224x224 image into compact frequency-domain tokens, achieving up to 128x reduction in off-chip data volume with modest accuracy loss; (2) a LUT-based near-sensor hardware implementation that leverages fixed DCT coefficients for multiplier-free, energy- and area-efficient tokenization; and (3) a modified MIPI-based low-power communication architecture that further reduces transfer energy. FrequencyFormer serves as a drop-in replacement for standard ViT patch embedding and remains compatible with pretrained backbones across classification, detection, and segmentation tasks. The pipeline achieves 28.8 TOPS/W, reduces communication energy by 230x, and lowers total sensor-side energy by 2.22x, demonstrating frequency-domain tokenization as a scalable foundation for in-sensor ViT deployment.

Decision-making with deep learning-based time series forecasting requires not only accurate predictions but also actionable insights. However, current architectures do not inherently provide such information. Specifically, guidance is needed on how current conditions must be modified to shift from a predicted outcome to a desired future scenario. Counterfactual explanations provide a natural framework for this task, as they represent minimal input changes that alter the model's prediction, indicating when and how intervention is required. Existing approaches rely on instance-wise optimization, leading to inconsistency across instances, high computational costs, and limited applicability in real-time settings. To address these limitations, we reformulate counterfactual generation for time series forecasting as the problem of learning a globally consistent intervention strategy, allowing counterfactuals to be generated through a single shared function. We propose Counterfactual Time Series Explanations (ConTex), a model-agnostic, decomposed architecture comprising a temporal context encoder and a conditional encoder, followed by two heads that capture interventions in terms of temporal relevance and modification strength. This structure overcomes the instability and inconsistency of instance-based approaches by producing targeted, interpretable interventions across time and feature dimensions in a single forward pass, making it suitable for real-time applications. Across multiple forecasting architectures and benchmark datasets, ConTex achieves state-of-the-art validity while generating sparse counterfactuals that minimize the number of necessary interventions. Additionally, our approach reduces computational cost by at least 12-36x compared to instance-wise generation and supports real-time inference at approximately 0.007 seconds.

Continuous AI inference on resource-constrained edge hardware introduces deployment effects that are largely invisible to conventional benchmark evaluation, including temporal instability in streaming video, thermal throttling under sustained load, and workload-dependent performance variability. We present Edge-TSR, a deployment-oriented continuous edge inference system for sustained roadside perception on the NVIDIA Jetson Orin Nano. Edge-TSR integrates detection, tracking, fine-grained classification, and a lightweight track-aware temporal stabilization mechanism that improves streaming inference consistency with negligible computational overhead. Our central finding is that benchmark-centric evaluation systematically overstates deployed edge inference performance. Across three state-of-the-art baselines, we observe consistent 20-30% relative degradation when transitioning from static-image evaluation to real-world streaming deployment. Edge-TSR addresses this gap through temporal inference stabilization, recovering up to 10.16% classification accuracy over per-frame inference baselines while maintaining sustained real-time performance under continuous operation. We evaluate the complete system under diverse real-world deployment conditions, jointly characterizing inference quality, latency, throughput, and thermal behavior during long-duration operation. A 55-minute vehicular deployment over a 26 km route demonstrates sustained operation at 16.18 FPS within safe thermal limits on a single embedded device without cloud offload. Our findings show that deployment-aware evaluation and temporal inference stabilization are necessary components of continuously operating edge AI systems intended for real-world sensing deployments. We release a sample annotated streaming video evaluation dataset and full system implementation to support reproducible deployment-centric evaluation.

Embedded devices from wildlife monitoring stations to clinical wearables require local AI inference due to latency, communication, or privacy constraints. Optimizing models for heterogeneous microcontrollers (MCUs) requires simultaneously satisfying hard physical constraints on memory, power, and temperature while preserving accuracy, a multidimensional optimization that is today performed manually by experts. We ask whether an LLM agent can autonomously navigate this complex, multi-turn pipeline guided by real hardware feedback, and introduce a hardware-in-the-loop agent arena in which the agent iteratively refines both model and firmware -- compiling, flashing, and measuring on real hardware -- to enable closed-loop optimization. Frontier models, including Claude Opus 4.7 and Gemini 3.1 Pro, fail entirely without hardware feedback (0% deployment success), whereas our hardware-in-the-loop formulation achieves the first successful deployment within three iterations and can surpass human expert results within seven. This agentic co-optimization achieves 250x compression for vision models with <3.3% accuracy loss and 400x for audio with <6% Feature Error Rate loss, enabling battery-free operation on a commercial MCU via solar harvesting. We demonstrate practical impact in two real-world systems: an elk-detection camera trap (96.7% accuracy) and a phonetic-transcription wearable (8.44% FER) for child development research.

Spatio-temporal prediction supports radar/satellite nowcasting and city-scale traffic monitoring, but modern models are often too expensive for real-time deployment. This stems from a mismatch between dense computation and strong input-dependent redundancy (e.g., calm seas or clear skies). To enable automated, resource-aware architecture optimization in scalable media analysis, we propose Dyna-Pruner, an end-to-end framework for input-dependent co-pruning of data and model structure. A shared-importance synchronization mechanism generates coupled masks that prune redundant regions and their corresponding computational units (e.g., convolutional filters), yielding per-sample sparse sub-networks at inference time. Experiments on WeatherBench, SEVIR, and TaxiBJ show seamless integration with CNN, RNN, and Transformer backbones, reducing FLOPs by up to $70\%$ and achieving a $2.5\times$ speedup on NVIDIA Jetson AGX Orin with negligible accuracy loss ($<1\%$).

Real-time fire classification systems require models that are simultaneously accurate, computationally efficient, and deployable on resource-constrained hardware. This work proposes \textbf{HumP-KD}, a Hybrid Uncertainty-aware Multi-stage Progressive Knowledge Distillation framework for efficient fire classification. Two datasets, FlameVision and Dataset-II, containing 8,600 and 31,309 images, are used. Various CNN and transformer baselines are applied under standard preprocessing, online augmentation, Gaussian noise and motion blur robustness conditions. The proposed HumP-KD model distills knowledge from two frozen heterogeneous transformer teachers, Swin-Tiny and ViT-Base, along with their Meta-MLP ensemble, into a lightweight MobileViT-S student via three tightly integrated components. Hierarchical Progressive Knowledge Distillation employs a Hierarchical Feature Builder. It generates a fused spatial attention mask to guide distillation toward discriminative regions selectively. Multi-Stage Knowledge Distillation progressively activates three distillation stages across training. On Dataset-II, HumP-KD achieves a mean F1 score of $0.9876 \pm 0.0063$ across 10 independent trials, significantly outperforming the MobileViT-S baseline trained without distillation ($0.9537 \pm 0.0351$), with statistical significance confirmed by both independent t-test ($p = 0.0195$) and Wilcoxon signed-rank test ($W = 1$, $p = 0.0039$). The proposed method also demonstrates strong generalization across datasets and robustness under degraded visual conditions. The student model retains only 4.94M parameters and 19.01Mb model size, representing a $5.7\times$ parameter reduction over Swin-Tiny and a $17.5\times$ reduction over ViT-Base, while achieving 37.72 CPU FPS, making it suitable for real-time deployment.

Real-time execution, enabled by asynchronous inference that ensures both smooth action trajectories and fast reactivity, is critical for realistic deployments of large-scale Vision-Language-Action models. However, recent work on real-time execution primarily focuses on variants of diffusion policies, even though it is more critical for autoregressive policies given their slower rollout speed in synchronous inference. In contrast, we demonstrate that autoregressive policies can achieve real-time execution by adjusting the tokenization horizon and applying constrained decoding, thereby guaranteeing strict latency bounds that enable multi-trajectory decoding to maximize performance. Across simulated and real-world environments, we find that the autoregressive policy consistently outperforms its equivalent-level flow-matching policy counterpart while achieving significantly improved task completion speeds from synchronous inference. Coupled with the inherent advantages of autoregressive policies, such as faster convergence and better generalizability in instruction-following, these results confirm that autoregressive policies can remain a competitive policy type supporting real-time execution.

The need for detecting and sorting batteries is drastically increasing for many applications. This study proves the potential of transfer learning in predicting whether the image contains a battery or not, the location and identifying three types of batteries, namely: prismatic, pouch, and cylindrical Lithium-Ion Batteries (LIB). Particularly, it focuses on the transfer learning method in two applications: Training a large-scale dataset to detect electronic devices using a pre-trained YOLOv5m, then using these latter trained weights to detect and classify the batteries. The precision of battery detection achieves 94%, which outperforms the pretrained YOLOv5m weights with 5%, in 22 ms inference time.

Diffusion-based lip synchronization models achieve strong visual quality and audio-visual alignment, but full-sequence bidirectional attention and many denoising steps make them impractical for real-time inference. We present Lip Forcing, to our knowledge the first autoregressive diffusion method for video-to-video (V2V) lip synchronization, which distills a 14B audio-conditioned bidirectional video diffusion teacher into causal students. At inference, the students generate each chunk in only two denoising steps without inference-time CFG, enabling real-time lip synchronization. A lip-sync-specific teacher-trajectory analysis reveals a CFG fidelity-sync tradeoff: no-CFG predictions favor reference fidelity, whereas CFG-guided predictions favor synchronization within a mid-trajectory band. Lip Forcing translates this finding into three analysis-derived components: Sync-Window DMD, a two-step inference schedule, and a SyncNet-based reward. We validate Lip Forcing at two student scales, both distilled from the 14B teacher. The 1.3B student crosses into real-time streaming at 31 FPS, $17.6\times$ faster than its same-scale bidirectional model. The 14B student, the largest diffusion model reported for V2V lip synchronization, runs $39.8\times$ faster than its teacher at comparable reference fidelity. Time-to-first-frame is sub-millisecond at both scales, far below every diffusion baseline.

Accurate calibration of particle-based simulators is crucial for robotic earthwork simulation, but analytical calibration is challenging due to this task's highly nonlinear particle dynamics and the black-box nature of conventional simulators. Although simulation-based inference (SBI) can estimate posterior distributions over simulation parameters solely from forward simulations, applying SBI directly to high-fidelity (HF) particle simulators is often computationally prohibitive. Low-fidelity (LF) simulators with coarser particles can reduce this cost, but changes in particle size and particle count shift the parameter values needed to reproduce the same observation, producing biased LF posteriors. We propose Bridged SBI, which leverages a biased but informative LF posterior to guide HF inference. This method first uses inexpensive LF simulations to identify a coarse high-density parameter region, and then it learns a local residual bridge to transport LF posterior samples toward HF-consistent regions by correcting the LF--HF discrepancy. We analyze how sequential multi-fidelity SBI (Naive-MF) can suffer from LF-induced posterior miscoverage when it directly relies on the LF posterior without discrepancy correction. We then show that Bridged SBI is designed to alleviate this issue by explicitly modeling the LF--HF discrepancy through residual correction. Experiments on both sim-to-sim particle-parameter calibration and real-to-sim calibration with real soil observation show that Bridged SBI produces more accurate and reliable HF posteriors than HF-only SBI or the Naive-MF baseline, especially under limited HF simulation costs.

Deploying vision--language--action (VLA) models on robots requires adapting model-specific inference pipelines to heterogeneous processors and limited onboard memory. We present vla.cpp, a unified C++ inference runtime for eleven VLA models, with no PyTorch dependency for model execution. The runtime shares model loading, tensor execution, and serving while retaining architecture-specific attention, conditioning, and action heads. Iterative policies reuse observation-dependent computation across solver steps, while regression policies predict actions directly. We evaluate task success on LIBERO-Object and profile supported configurations on NVIDIA, Apple, and Intel hardware. BitVLA completes 200/200 LIBERO-Object episodes on an 8,GB Jetson Orin Nano. A ternary tensor-core kernel accelerates its client inference by $4.0$--$4.6\times$ over the CUDA-core baseline on RTX 3060 and AGX Orin. A SmolVLA case study links positional-index precision to gripper commands and task success, showing why fixed-input numerical checks should accompany rollout evaluation. Deployments on UR10e and ALOHA demonstrate physical robot integration; delay and execution-horizon studies characterize synchronous chunked control. The results demonstrate a common deployment path across VLA architectures and hardware, with numerical validation and control settings guiding deployment alongside inference efficiency.

Language-guided UAV agents must execute long-horizon semantic instructions while producing smooth, physically feasible continuous flight commands, yet existing Vision-Language Navigation (VLN) benchmarks typically use discrete or coarse actions and existing UAV Vision-Language-Action (VLA) tasks focus on short, atomic maneuvers. To address this gap in UAV task settings, we introduce \textbf{FLIGHT}, a \textbf{F}ine-grained \textbf{L}ong-horizon \textbf{I}nstruction-\textbf{G}uided benchmark for \textbf{H}ybrid UAV navigation and reasoning \textbf{T}asks, which combines multi-stage instructions with dense 6-DoF trajectory annotations across two dataset splits: Fine-grained VLN and Long-horizon Flow. To endow the UAV agent with the capability of real-time in-flight reasoning over task execution status and mission planning, while simultaneously accommodating high-frequency, real-time precise control, we further propose \textbf{FLIGHT VLA}, an asynchronous architecture that decouples a low-frequency Streaming Pilot Vision-Language Model (VLM) for task-state reasoning from a high-frequency diffusion action model for continuous control, supervised by explicit \textbf{Pilot Reasoning} texts that summarize the current flight state and anticipate the next subgoal. In closed-loop evaluation, FLIGHT VLA consistently surpasses representative VLN and VLA baselines on our FLIGHT benchmarks, achieving stronger multi-stage completion, subgoal adherence, and terminal control. Its trained Streaming Pilot Reasoning VLM further improves UAV video reasoning, validating the effectiveness of our design.

Designing neural architectures for time-series forecasting and anomaly detection remains a resource-intensive task that often requires substantial domain expertise. Traditional Automated Machine Learning (AutoML) systems typically rely on static, predefined search spaces, limiting their ability to adapt to diverse data characteristics. We present GenAutoML, an agentic framework that leverages Large Language Models (LLMs) as neural architects to bridge natural-language requirements and executable PyTorch implementations. The framework incorporates a Sandboxed Reflection Loop for autonomous code refinement and a Signature-Aware Runtime that enforces architectural consistency and execution safety. To improve robustness under non-stationary conditions, we further introduce a Dynamic Reversible Instance Normalization (Dyn-RevIN) wrapper. Experiments on the ETTh1, ETTm1, and Weather benchmarks demonstrate that GenAutoML can dynamically generate task-specific neural architectures tailored to dataset characteristics. Among the generated models, WaveInterferenceNet achieves inference latency below 0.01 ms per sample while maintaining competitive predictive performance. By emphasizing computational efficiency, architectural adaptability, and stable optimization behavior, GenAutoML enables the creation of ultra-lightweight neural networks suitable for resource-constrained and latency-sensitive Edge AI deployments.

The development of multilingual Alzheimer's Disease Dementia (AD) detection models presents significant challenges due to the resource-intensive and time-consuming nature of language-specific model training. We propose a novel solution using cross-language training to detect AD in languages beyond those used for model training. This study investigates multilingual deep learning models for detecting AD across different languages and cognitive impairment levels. Using datasets in English, Chinese, Arabic, and Hindi, we developed transformer-based models for binary AD classification. Our approach achieved F1 scores of 82% across all languages, demonstrating strong cross-linguistic generalization. The rapid inference time (0.5 seconds) supports potential real-time screening applications, while consistent performance across languages indicates feasibility for global deployment.

Audio is an inherently interactive modality, yet today's Large Audio Language Models (LALMs) are offline, and streaming audio models each handle only a single task such as streaming ASR or voice chatting. It is time to unify them into one online LALM: a model that, through an always-on perceive-decide-respond loop, listens to sound, environment, and instructions in real time and reacts on the fly. We formalize this regime as the Audio Interaction Model, and realize it with Audio-Interaction, a unified streaming model that retains offline task execution while adding online general audio instruction following, from dialogue to full voice chatting, deciding when to respond from the semantics of the stream. To enable this, we propose SoundFlow, a framework that instantiates the perceive-decide-respond loop end to end, from data to training to deployment, through streaming-native data construction, comprehension-aware training, and asynchronous low-latency inference for stable real-time interaction. We further construct StreamAudio-2M, a 2.6M-item streaming corpus spanning 7 fundamental abilities and 28 sub-tasks, and Proactive-Sound-Bench for evaluating proactive audio intervention. Across 8 benchmarks, Audio-Interaction preserves competitive performance on mainstream audio tasks while unlocking capabilities inaccessible to offline LALMs, including real-time ASR, streaming audio instruction following, and proactive help.

Real-time monitoring of the temperature distribution within components and sub-structures is a challenging topic in many systems due to restrictions on feasible sensor locations. While machine learning (ML) proves a versatile tool in many applications, its adoption for high-resolution thermal monitoring is hindered by the availability of high-quality datasets for training. In this work, we propose a novel approach for generating datasets for industrial applications based on randomized physics-based simulations. We demonstrate the approach in a proof-of-concept hardware setup: A neural network (NN) trained only on such a synthetic dataset, is used to reconstruct the internal temperature field from sparse sensors embedded in the hardware. The NN-based reconstructions do not only outperform Kriging in robustness but also enable real-time inference, making the method suitable for online monitoring of otherwise unobservable thermal states.

The explosive growth of Text-to-Image (T2I) models, from large-scale versions to lightweight, real-time ones, now faces diminishing marginal returns from single-model scaling. Agentic T2I methods emerged to alleviate this bottleneck by using multiple models. However, existing agentic T2I methods suffer from three key challenges: reliance on expensive handcrafted priors or human annotations, rigid single-path decision mechanisms, and a neglect of inference efficiency. To address these challenges, we introduce OctoT2I, a novel agentic framework that reformulates the T2I task as a joint optimization of generation quality and inference efficiency. OctoT2I implements a stateful, multi-round routing strategy that adaptively selects the most suitable tool based on its knowledge and memory. This strategy is enabled by a knowledge base built from scratch by our novel Self-Evolving Mechanism. This mechanism, which requires no human supervision, first autonomously defines foundational Conceptual Dimensions (eg, style, color, count) and then intelligently explores their combinations via an iterative" Propose--Solve--Evaluate--Learn"(PSEL) loop. The PSEL loop efficiently discovers each tool's capability frontier, driving continuous improvement without external guidance. Extensive experiments demonstrate that OctoT2I achieves competitive performance (0.96) on GenEval while delivering a 90.3% inference speedup and a 56.6% energy-efficiency gain over the leading baseline (Flow-GRPO), striking an exceptional balance between performance and efficiency. Code and models will be made available.

While diffusion-based policies have impressive performance and expressivity, their long offline training slows down the data collection and policy deployment loop. We introduce Closed-Form Diffusion Policies, a class of training-free diffusion-based policies for imitation learning using the closed-form score derived from the demonstration dataset. We deploy CFDP with real-time inference with a mobile CPU in hardware experiments, showing it can successfully perform imitation directly from the dataset in milliseconds and with faster inference than neural diffusion policies. In experiments on imitation learning benchmarks, we show that CFDP is competitive against neural baselines that require hours of training, providing a favorable tradeoff between training time and performance. Finally, we show how closed-form diffusion policies act as a composable primitive that enables data-driven inference-time editing of pre-trained neural diffusion policies, including policy guidance and novel demonstration augmentation.

Recent work has begun to equip vision-language-action (VLA) policies with explicit intermediate reasoning. In embodied control, however, textual chain-of-thought is a poor fit: irrelevant or weakly textual information can interfere with action prediction, while autoregressive text decoding adds too much latency for real-time closed-loop execution. We present VISUALTHINK-VLA, a visual intermediate-reasoning framework for accurate, low-latency VLA policies. Our bootstrapping philosophy is to guide action with effective visual thinking: VISUALTHINK-VLA bootstraps action prediction through a compact visual-evidence interface that preserves spatial precision while avoiding decoding overhead. Besides, to further improve performance and efficiency, VISUALTHINK-VLA adopts a tailored selective routing mechanism to learn the visual evidence tokens, enabling low-latency inference while preserving high-capacity specialization. We also introduce VisualEvidence-Kit, a supervision-and-audit resource centered on a VisualEvidence-Agent that constructs a 754.7k VLA instructions VisualEvidence-Set for route supervision and counterfactual faithfulness tests. Across multiple benchmarks and real-robot evaluation, VISUALTHINK-VLA achieves the highest success rate on most benchmarks while reducing the multi-second latency of reasoning-augmented baselines to the sub-second regime. For example, on BridgeData V2, it reduces step latency from 8.377,s with ECoT to 0.367,s, achieving a 22.8 times speedup.

Feed-forward networks (FFNs) dominate the parameter count and computation of modern language models, yet existing pruning methods often struggle to convert sparsity into hardware-friendly inference efficiency gains. We introduce \textbf{PrunePath}, a budget-adaptive structured sparsification framework for FFN layers. Built on MoEfication, PrunePath replaces independent expert-wise thresholding with a softmax-normalized routing distribution and activates important experts under a cumulative-mass threshold. This formulation imposes a token-level probability budget, enabling adaptive expert counts and a direct inference-time sparsity knob from a single checkpoint. Across NLU, NLG, and instruction-tuning evaluations, PrunePath achieves a favorable sparsity--performance trade-off compared with existing static pruning and MoEfication-based methods. We further implement Triton kernels for KV-cache decoding to translate the resulting structured sparsity into practical memory savings and measurable decoding-speed improvements. These results demonstrate the superior performance of PrunePath for building highly sparse, deployment-friendly large language models.

Mobile graphical user interface (GUI) agents enable AI models to autonomously operate smartphones on behalf of users. However, most existing systems focus primarily on optimizing task accuracy and rely on cloud-hosted models for inference, which introduces privacy concerns and network-dependent latency. As a result, fully on-device deployment of mobile GUI agents remains underexplored. We propose MobileExplorer, a new framework that accelerates on-device inference for vision-based mobile GUI agents via online exploration. The key idea is to exploit the long per-step reasoning time of vision-language models (VLMs) by performing lightweight, parallel exploration of UI elements. During model inference, the agent proactively probes semantically relevant UI elements and records these exploration traces as structured memory. To ensure reliable execution in live mobile environments, we design a two-level rollback mechanism that robustly restores the initial UI state when a fast but naive backtracking strategy fails. The collected exploration traces are then summarized into concise contextual hints and injected into the prompt to enhance the subsequent reasoning step. We evaluate MobileExplorer on multiple off-the-shelf devices using the AndroidWorld benchmark, as well as newly designed, more complex tasks and dynamic on-device environments. MobileExplorer reduces the average number of reasoning steps and end-to-end latency by 23%, while maintaining or improving task success rates by up to 5%. A video demonstration of MobileExplorer performance in the real world is available at https://youtu.be/thK7MJmdlvM .

Large-batch Contrastive Learning (CL), the foundation of modern representation learning, is fundamentally incompatible with the volatile resource constraints of edge devices. This conflict creates a dilemma: small on-device batches degrade model fidelity, while offloading to the cloud incurs unacceptable latency and bandwidth costs. Existing solutions often resort to static model compression, which fails to adapt to the runtime volatility of edge environments. To bridge this gap, we present StreamSplit, a novel framework that makes streaming CL practical across heterogeneous ARM client platforms. StreamSplit resolves the conflict between the continuous nature of ambient audio and the discrete batch requirements of models like CLAP and COLA. We introduce: (1) A distribution-based streaming framework that decouples representation quality from local batch size, using a tractable Hybrid Loss to maintain fidelity despite sparse updates; and (2) An Uncertainty-Guided Adaptive Splitter that uses a lightweight Reinforcement Learning (RL) policy to dynamically partition computation. Uniquely, this policy integrates real-time resource monitoring with embedding ambiguity to optimize the accuracy-latency trade-off on the fly. We evaluate StreamSplit on diverse hardware, from the resource-constrained Raspberry Pi 4 to the high-performance Apple M2. Results demonstrate that StreamSplit reduces per-sample latency by up to 4.7x and cuts bandwidth by 77.1% and energy by 52.3% compared to server-centric baselines. Crucially, it maintains accuracy within 2.2% of server-centric models, proving that adaptive, distributed learning is a viable path for the modern edge ecosystem.

Robots deployed in dynamic environments must contend with environment-driven changes that reshape computation at runtime: new tasks may appear, precedence relations can shift, and overall workload structure evolves, all of which degrade performance, especially when multi-task inference is required under tight resource and real-time budgets. We present RED, a real-time scheduling framework for multi-task deep neural network workloads on resource-constrained robotic platforms that adapts to Robotic Environmental Dynamics (RED) while preserving end-to-end timing guarantees under modeling assumptions. The core of RED is a deadline-aware scheduler that assigns intermediate sub-deadlines, allowing it to accommodate evolving computation graphs and asynchronous inference induced by unpredictable conditions. The framework also supports flexible deployment of MIMONet (multi-input multi-output neural networks), commonly used in multi-tasking robots to alleviate memory pressure through weight sharing. RED explicitly leverages this shared-parameter property via a workload refinement and graph-reconstruction procedure that aligns MIMONet structure with schedulability requirements, improving compatibility and efficiency. We implement RED on NVIDIA Jetson family platforms and on an Apple M-series MacBook and evaluate it on navigation-oriented workloads representative of real robotic scenarios. Experiments show consistent gains over existing methods in throughput, deadline satisfaction, robustness to interference, adaptability, and runtime overhead.

Real-time cognitive load assessment from eye-tracking signals could enable adaptive human-centered AI in safety-critical applications such as driver vigilance monitoring or automated flight deck assistance, yet two challenges persist: handling frequent data missingness from blinks and tracking failures, and efficiently modeling long-range temporal dependencies. We propose MambaGaze (Bi-Mamba), a framework that addresses these challenges through (1)~XMD encoding, which augments raw features with observation masks and time-deltas to explicitly model data uncertainty, and (2)~bidirectional Mamba-2, which captures temporal dependencies with linear computational complexity. Experiments on CLARE and CL-Drive datasets under leave-one-subject-out evaluation show that MambaGaze achieves 77.1% accuracy and 59.2% macro-F1 on CLARE, and 69.4% accuracy and 51.5% macro-F1 on CL-Drive, attaining the highest average LOSO macro-F1 (55.3%) across all ten compared models. Input-stream ablation indicates that log-scaled time-deltas are the strongest single channel in our setting, and combining all three XMD streams provides consistent gains of 5--20,pp macro-F1. Edge deployment benchmarks on three NVIDIA Jetson Orin platforms show real-time inference at 27--36,FPS with power consumption below 6.6,W, supporting feasibility for embedded cognitive load monitoring.

Adapting pretrained diffusion models to downstream objectives such as inverse problems often requires expensive test-time guidance or optimization. We propose a principled framework for generating high-quality reward-aligned samples at substantially reduced inference cost. Our approach formulates test-time adaptation as a hierarchical variational model, where control is amortized into a lightweight yet expressive stochastic policy. This formulation naturally supports few-step diffusion sampling: large step sizes enable fast inference, while the learned policy maintains sample quality by providing structured per-step control. The resulting fully amortized sampler achieves a strong quality--speed tradeoff, matching or exceeding recent test-time scaling baselines while requiring significantly less compute. For example, on 4x super-resolution, our method achieves better perceptual quality with more than 5x faster inference compared to the best-performing baseline. We further extend our approach to a semi-amortized regime that combines cheap amortized proposals with limited test-time optimization, achieving state-of-the-art perceptual quality across several challenging inverse problems.

Vision-Language-Action (VLA) policies increasingly rely on asynchronous inference to hide large-model latency behind ongoing robot motion. While this avoids the stop-and-go behavior of synchronous action-chunk execution, it creates a prediction-execution mismatch: the next chunk is computed from a stale observation at inference start but executed only after the robot and scene have evolved. As a result, actions that fit the prediction-time state can become misaligned with the execution-time state. Existing runtime repair, behavior-cloning, and preference-alignment approaches do not directly teach the policy to resolve this stale-input mismatch. We propose DEFLECT, an offline post-training framework for delay-robust asynchronous VLAs. DEFLECT converts latency-induced mismatch into counterfactual preference supervision: a frozen reference VLA generates a preferred chunk from the future execution-time observation and a rejected chunk from the stale prediction-time observation. The trainable policy scores both chunks under the same deployment-time input, learning to favor execution-time-aligned actions while a supervised fine-tuning anchor preserves the expert action manifold. DEFLECT requires no human preference labels, reward models, online robot rollouts, architectural changes, or additional inference-time computation. Across Kinetix, LIBERO, and three real-robot tasks, DEFLECT improves delay robustness over strong asynchronous VLA baselines, raising high-latency success by up to 6.4 percentage points and achieving a 4.6 percentage-point gain at the longest delay on a real-scale VLA.

Large language model (LLM) agents often rely on long sequences of low-level textual actions, resulting in large effective decision horizons and high inference cost. While prior work has focused on improving inference efficiency through system-level optimizations or prompt engineering, we argue that a key bottleneck lies in the representation of the action space itself. We propose Latent Action Reparameterization (LAR), a framework that learns a compact latent action space in which each latent action corresponds to a multi-step semantic behavior. By reparameterizing agent actions into latent units, LAR enables decision making over a shorter effective horizon while preserving the expressiveness of the original action space. Unlike hand-crafted macros or hierarchical controllers, latent actions are learned from agent trajectories and integrated directly into the model, allowing both planning and execution to operate over abstract action representations. Across a range of LLM-based agent benchmarks, LAR significantly reduces the effective action horizon and improves inference efficiency under fixed compute budgets. As a consequence, our approach achieves substantial reductions in action tokens and corresponding wall-clock inference time, while maintaining or improving task success rates. These results suggest that action representation learning is a critical and underexplored factor in scaling efficient LLM agent inference, complementary to advances in model architecture and hardware.

The rapid growth of online video platforms and AI-generated content has made reliable video guardrails a key challenge for safety and real-world deployment. While most videos can be screened through fast pattern recognition, a small subset requires deeper reasoning over temporally complex content and nuanced policy constraints. Existing approaches typically rely on large vision-language models applied uniformly across all inputs, resulting in high inference costs and inefficient allocation of computation. We propose SafeLens, a video guardrail framework that introduces a fast-and-slow inference architecture for efficient and accurate content moderation with variable computational cost across inputs. Additionally, we construct a high-quality dataset by applying influence-guided filtering to the SafeWatch Dataset, retaining only 2.4% of the original data. To further address limitations of training-time scaling, we enable test-time reasoning by augmenting the filtered data with structured Chain-of-Thought traces. Across real-world and AI-generated video benchmarks, SafeLens achieves state-of-the-art performance, outperforming strong open-source video guardrails (e.g., SafeWatch-8B, OmniGuard-7B) and closed-source models (e.g., GPT-5.4, Gemini-3.1-pro) while significantly reducing inference cost, demonstrating that efficient design serves to be more effective than scaling data or model size alone.

The Diffusion Transformer (DiT) architecture is the state-of-the-art paradigm for high-fidelity image generation, underpinning models like Stable Diffusion-3 and FLUX.1. However, deploying these models on resource-constrained mobile devices entails prohibitive computational and memory overhead. While efficiency-driven approaches like Linear-DiT and static pruning alleviate bottlenecks, they often incur quality degradation. Unlike cloud environments, mobile constraints require a single-model paradigm that dynamically balances fidelity and latency. We introduce ElasticDiT, which achieves this dynamic trade-off by adjusting spatial compression ratios and DiT block depths. By integrating Shift Sparse Block Attention (SSBA) and a Tiny DWT-Distilled VAE (T-DVAE), ElasticDiT reduces inference latency and memory footprint while maintaining image quality. Experiments confirm that ElasticDiT effectively covers a wide range of fidelity-latency trade-offs within a single set of parameters. By jointly adjusting compression and depth, a single ElasticDiT model can be reconfigured on-the-fly to outperform task-specific baselines. Specifically, our flex lite variant achieves an HPS of 32.87, surpassing the Flux model, while maintaining competitive quality at 84.16 percent average sparsity through SSBA. Furthermore, the plug-and-play T-DVAE provides SD3-level reconstruction with only 1/8x the computational cost of standard VAEs, and Flow-GRPO boosts semantic alignment (GenEval: 66.93 to 73.62). These results demonstrate that ElasticDiT offers a versatile, hardware-adaptive solution that eliminates the need for multiple specialized models, providing a promising path for future high-resolution image generation on mobile devices.

Large Language Diffusion Models (LLDMs) are emerging as an alternative to autoregressive models, offering faster inference through higher parallelism. Similar to autoregressive LLMs, they remain prone to hallucinations, making reliable uncertainty quantification (UQ) crucial for safe deployment. However, existing UQ methods are fundamentally misaligned with this new paradigm: they assume autoregressive factorization or use expensive repeated sampling, negating the efficiency of LLDMs. In this work, we present the first systematic study of UQ for LLDMs and propose lightweight, zero-shot uncertainty signals derived from the iterative denoising process, leveraging intermediate generations, token remasking dynamics, and denoising complexity. We further adapt a state-of-the-art UQ method to LLDMs by combining masked diffusion likelihoods with trajectory-based semantic dissimilarity. We prove that expected trajectory dissimilarity lower bounds the masked diffusion training objective, which motivates its usage as an uncertainty score. Comprehensive experiments across three tasks, eight datasets, and two models show that our method achieves a great cost-performance trade-off: it approaches the strongest sampling-based baselines while incurring up to 100x lower computational overhead. Our work demonstrates that LLDMs can deliver both fast inference and reliable hallucination detection simultaneously.

Conventional transformer inference engines are request-driven, paying an O(n) prefill cost on every query. In streaming workloads, where data arrives continuously and queries probe an ever-growing context, this cost is prohibitive. We introduce a data-driven computational model centred on stateful sessions: a persistent KV cache advanced incrementally as new data arrives, so prefill is moved off the critical path and query latency becomes O(|q|), independent of accumulated context size. Building on this, Flash Queries reclaim idle GPU cycles between data arrivals to pre-evaluate registered questions and return cached answers before the user asks, a pattern that is structurally impossible in stateless engines because they discard intermediate state between requests. A multi-tenant continuous-batching scheduler with cell-budget admission and prefix-aware grouped prefill lets dozens of stateful sessions coexist on a single GPU while preserving full quadratic self-attention. On streaming market-data benchmarks the reference implementation achieves up to 5.9x speedup over conventional inference engines (vLLM, SGLang, TensorRT-LLM, llama.cpp), holding query latency constant as accumulated context grows.

Edge deployment of Vision-Language Models (VLMs) faces a tradeoff between latency and accuracy: cloud execution provides high-quality predictions but incurs communication delay and energy cost, while edge-only execution is faster but less accurate due to limited model capacity. This trade-off is further complicated by heterogeneity in image quality and reasoning complexity, making static placement suboptimal. We present INAR-VL, a lightweight edge-cloud routing system for multimodal inference in a two-tier deployment. INAR-VL maintains complementary VLMs across edge and cloud and uses lightweight image and text complexity signals to guide routing and model selection, executing simple queries locally while offloading complex ones when beneficial. Evaluation on visual question answering shows that INAR-VL executes 36% of requests on the edge, reduces latency by 24%, lowers energy by 26%, and preserves 97% of cloud-level accuracy.

The increasing integration of sensors in autonomous maritime navigation has led to large-scale multimodal datasets, raising challenges in achieving efficient real-time perception. In such systems, object detection and trajectory perception of nearby vessels are tightly coupled, particularly in dynamic environments such as maritime navigation. However, the efficiency of object detection models during inference remains an often-overlooked aspect. To this end, we build upon an existing object detection framework by incorporating GPS trajectory data into the inference process to enable input-adaptive computation. Specifically, we introduce an early-exit mechanism in a YOLOv8-based detector that incorporates motion cues - such as inter-vessel distances. Frames of vessels that are separated by short distances, converging with high speed, are processed using the full model, while only a subset of the network's architecture is activated otherwise. The difficulty degree (or scene complexity) of a frame or set of frames per second is evaluated by leveraging inter-object distance and the rate at which the distance between them decreases. Experimental results demonstrate that this strategy maintains satisfactory detection performance while significantly reducing inference time and computational cost, thus enabling a flexible trade-off between accuracy and efficiency compared to full-model inference.

SNNs promise energy-efficient and low-latency inference, but their performance still trails that of ANNs. ANN-to-SNN knowledge distillation helps narrow this gap, yet the original training data are often unavailable in practical deployment settings. Existing data-free knowledge distillation (DFKD) methods synthesize surrogate data by matching teacher-side priors, especially BN statistics, but these ANN-oriented constraints mainly regularize mean and variance and therefore remain under-constrained for SNN students whose responses depend on threshold-crossing dynamics. In this paper, we propose Spike Tail-Aware Relational Synthesis (STARS), a plug-and-play method for ANN-to-SNN DFKD that augments standard BN-guided synthesis with two complementary objectives: Relational Consistency Alignment, which preserves cross-sample relational consistency between teacher and student, and Tail-Aware Regularization, which regularizes threshold-relevant tail probabilities through soft exceedance over teacher-derived thresholds. Together, these objectives generate synthetic batches that remain teacher-valid while becoming more informative for SNN students. Experiments on CIFAR-10, CIFAR-100, and Tiny-ImageNet across multiple ANN-SNN pairs show that our method consistently improves conventional DFKD baselines and even surpasses several KD methods, with gains of up to 4.6% on CIFAR-10 and 6.7% on CIFAR-100, highlighting the importance of complementing BN matching with relational and tail-aware constraints in SNN-oriented DFKD.

Diffusion Language Models (DLMs) have recently emerged as a promising alternative to autoregressive language models, offering stronger global awareness and highly parallel generation. However, post-training DLMs with standard Negative Evidence Lower Bound (NELBO)-based supervised fine-tuning remains inefficient: training reconstructs randomly masked tokens in a single step, whereas inference follows a confidence-guided, multi-step easy-to-hard denoising trajectory. Recent trajectory-based self-distillation methods exploit such inference trajectories mainly for sampling-step compression and acceleration, often improving decoding efficiency without substantially enhancing the model's underlying capability, and may even degrade performance under full diffusion decoding. In this work, we ask whether self-distilled trajectories can be used not merely for faster inference, but for genuine knowledge acquisition. Although these trajectories lie on the pretrained DLM's own distributional manifold and thus offer a potentially lower optimization barrier, we find that naively fine-tuning on them with standard NELBO objectives yields only marginal gains. To address this limitation, we propose \textbf{T}rajectory-\textbf{A}ligned optimization via \textbf{Bo}ltzmann \textbf{M}odeling (\textbf{TABOM}), a self-distilled trajectory-based post-training framework that aligns training with the easy-to-hard structure of inference. TABOM models the inference unmasking preference as a Boltzmann distribution over predictive entropies and derives a tractable pairwise ranking objective to align the model's certainty ordering with the observed decoding trajectory. Empirically, TABOM achieves substantial gains in new domains, expands the effective knowledge boundary of DLMs, and significantly mitigates catastrophic forgetting compared with standard SFT.

End-to-end Vision-Language-Action (VLA) models for autonomous driving unify perception, reasoning, and control in a single neural network, achieving strong driving performance but requiring 20-60GB of GPU memory-far exceeding the 12-16GB available on commodity GPUs. We present a framework, which enables memory-efficient VLA inference on VRAM-constrained GPUs through system-level optimization alone, without model modification. Our work proceeds in three stages: (1) Sequential Demand Layering reduces VRAM usage from model-level to layer-level granularity; (2) Pipelined Demand Layering hides parameter transfer time within layer execution time via transfer--compute overlap; and (3) a GPU-Resident Layer Decision Policy, informed by per-module residency benefit analysis, eliminates the residual transfer overhead that pipelining cannot hide. We further propose a performance prediction model that determines the optimal configuration-both the number and placement of resident layers-from a single profiling run with less than 1.3% prediction error across all configurations. Applied to NVIDIA's Alpamayo-R1-10B (21.52GB) on an RTX 5070Ti (16GB), our work achieves up to 3.55x speedup over Accelerate offloading while maintaining full BF16 precision.

The Vision Transformer (ViT) achieves remarkable accuracy across visual tasks but remains computationally expensive for edge deployment. This paper presents MicroViTv2, a lightweight Vision Transformer optimized for real-device efficiency. Built upon the original MicroViT, the proposed model is designed based on reparameterized design, specifically Reparameterized Patch Embedding (RepEmbed) and Reparameterized Depth-Wise convolution mixer (RepDW) for faster inference, and introduces the Single Depth-Wise Transposed Attention (SDTA) to capture long-range dependencies with minimal redundancy. Despite slightly higher FLOPs, MicroViTv2 improves accuracy up to 0.5% compared to its predecessor and surpassing MobileViTv2, EdgeNeXt, and EfficientViT while maintaining fast inference and high energy efficiency on Jetson AGX Orin. Experiments on ImageNet-1K and COCO demonstrate that hardware-aware design and structural re-parameterization are key to achieving high accuracy and low energy consumption, validating the need to evaluate efficiency beyond FLOPs. Code is available at https://github.com/novendrastywn/MicroViT.

Recent advances in artificial intelligence (AI) have enabled effective perception and language models for robots, but their deployment remains computationally expensive, increasing latency and energy use. This work presents the Open Robotics Inference and Control Framework (ORICF), a modular, declarative, and model-agnostic platform for composing multimodal robotic inference pipelines. ORICF integrates input/output (I/O) adapters, pluggable inference back ends, and post-processing logic, while lightweight YAML specifications allow models, hardware targets, and data channels to be changed without code modification. The framework also supports edge offloading, i.e., executing inference on nearby external computers instead of onboard the robot. ORICF is evaluated on a mobile robot that answers spoken queries about people detected in its camera stream by combining automatic speech recognition (ASR), a large language model (LLM), and a convolutional neural network (CNN) detector through Robot Operating System 2 (ROS2). Compared with onboard execution, ORICF-based edge deployment reduces robot-side compute utilization by up to 83.16% and estimated energy consumption by 65.8%, while preserving modularity and reproducibility.

Low-cost air quality sensors are becoming ubiquitous in our daily lives as public awareness of air pollution continues to grow, and people take measures to monitor and improve the air they breathe indoors. Besides the standard operation of these sensors, fluctuations in environmental parameters can be leveraged to understand human behavior and activities in indoor spaces. Unlike traditional audio-visual, Radio Frequency, and inertial sensors, air quality sensors are easily scalable to a household, are privacy-preserving, and more economical. Such distributed sensor networks must jointly make decisions to monitor indoor occupants for downstream smart home and healthcare applications. However, due to low processing power, memory, and energy, they often struggle to maintain distributed data consensus and identify activity-affected sensor groups for accurate on-device inference. In this paper, we propose PoHAR framework that implements: (i) a conflict-free replicated data primitive for data sharing, (ii) a hierarchical clustering for ESP32 to detect activity-affected sensor groups with a self-supervised distance metric, and (iii) a leader-based group inference with off-the-shelf ML classifiers, enabling the sensor network to collaboratively detect hyperlocal indoor activities. Our extensive experiments demonstrated on-device activity detection, achieving 97.41% accuracy for indoor activity and 99.68% for cooking activity, using off-the-shelf ML models with latency below 34 microseconds.

Large language models (LLMs) are transforming society, powering applications from smartphone assistants to autonomous driving. Yet cloud-based LLM services alone cannot serve a growing class of applications, including those operating under intermittent connectivity, sub-second latency budgets, data-residency constraints, or sustained high-volume inference. On-device deployment is in turn constrained by limited computation and memory. No single endpoint can deliver high-quality service across this spectrum. This article focuses on collaborative intelligence, a paradigm in which multiple independent LLMs distributed across device and cloud endpoints collaborate at the task level through natural language or structured messages. Such collaboration strives for superior response quality under heterogeneous resource constraints spanning computation, memory, communication, and cost across network tiers. We present collaborative inference along two complementary and composable dimensions: vertical device-cloud collaboration and horizontal multi-agent collaboration, which can be combined into hybrid topologies in practice. We then examine learning to collaborate, addressing the training of routing policies and the development of cooperative capabilities among LLMs. Finally, we identify open research challenges including scaling under resource heterogeneity and trustworthy collaborative intelligence.

Object detection on heterogeneous edge devices must satisfy strict energy, latency, and memory constraints while still providing reliable perception for downstream autonomy. Existing energy-aware NAS methods often target limited deployment settings, while real energy remains difficult to optimize because it is highly device-dependent and costly to measure. We address these challenges with an energy-adaptive framework that combines an energy-aware XiResOFA search space, a two-stage energy estimator, and iterative search to identify a single energy-efficient base architecture. We then apply compound scaling to transform this base design into the XiYOLO family across deployment budgets, enabling interpretable accuracy-energy tradeoffs under sparse hardware measurements. Experiments on PascalVOC, COCO, and real-device deployment show that XiYOLO achieves a stronger energy-accuracy tradeoff than YOLO baselines. On PascalVOC, the medium XiYOLO model reaches 86.15 mAP50 while reducing energy relative to YOLOv12m by 20.6% on GPU and 35.9% on NPU. On COCO, XiYOLO reduces energy relative to YOLOv12 by up to 53.7% on GPU and 51.6% on NPU at the small scale. The proposed two-stage estimator also improves sample efficiency over a joint predictor under few-shot adaptation with only 2-20 target-device samples.

This paper presents CARMEN, a runtime-adaptive, CORDIC-accelerated multi-precision vector engine for resource-efficient deep learning inference. The key insight is that CORDIC iteration depth directly governs computational accuracy, enabling dynamic switching between approximate and accurate execution modes without hardware modification. The architecture integrates a low-resource iterative CORDIC-based MAC unit with a time-multiplexed multi-activation function block, supporting flexible 8/16-bit precision and high hardware utilization. ASIC implementation in 28 nm CMOS achieves up to 33% reduction in computation cycles and 21% power savings per MAC stage; a 256-PE configuration delivers 4.83 TOPS/mm2 compute density and 11.67 TOPS/W energy efficiency. FPGA deployment on PynqZ2 validates 154.6 ms latency at 0.43 W for real-time object detection.

Local execution of AI on edge devices is important for low latency and offline operation. However, deploying models on diverse hardware remains fragmented, often requiring model conversion or complete reimplementation outside the PyTorch ecosystem where the model was originally authored. We introduce ExecuTorch, a unified PyTorch-native deployment framework for edge AI. ExecuTorch enables seamless deployment of machine learning models across heterogeneous compute environments. It scales from embedded microcontrollers to complex system-on-chips (SoCs) with dedicated accelerators, powering devices ranging from wearables and smartphones to large compute clusters. ExecuTorch preserves PyTorch semantics while allowing customization, support for optimizations like quantization, and pluggable execution "backends". These features together enable fast experimentation, allowing researchers to validate deployment behavior entirely within PyTorch, bridging the gap between research and production.

Vision-Language-Action (VLA) models offer a promising path to generalist robot control, but their inference latency causes observation staleness when generated actions are executed asynchronously. Several methods have been proposed concurrently to mitigate this problem: inference-time inpainting (IT-RTC), training-time delay simulation (TT-RTC), future-state-aware conditioning (VLASH), and lightweight residual correction (A2C2). Each takes a fundamentally different approach, but they have so far been evaluated independently with different codebases, base policies, and protocols. We present a systematic comparison of these four methods under controlled conditions. We develop two unified codebases that integrate all methods with harmonized library and dataset versions, and we benchmark them on the Kinetix suite with MLPMixer policies and on the LIBERO manipulation benchmark with SmolVLA, sweeping inference delays up to $d=20$ control steps. A2C2's per-step residual correction is the most effective method on Kinetix, holding above 90% solve rate up to $d=8$, and also leads on LIBERO from $d=4$ onwards. IT-RTC is competitive at low delays but degrades sharply under long chunks ($H=30$) and high delays. TT-RTC is the most robust training-based method: stable across d_\max choices, generalizes beyond its training delay distribution, and adds zero inference overhead. VLASH exhibits a clear low-delay vs. high-delay trade-off governed by the fine-tuning delay range [0,d_\max]. Code is available at https://github.com/TheAyos/async-vla-inference

We propose a modular framework for hybrid image restoration that integrates transformer and state-space model (SSM) blocks with a focus on improving runtime efficiency on edge hardware. While transformers provide strong global modeling through self-attention, their attention kernels incur substantial latency on mobile devices, especially for high-resolution inputs. In contrast, SSMs such as Mamba offer lineartime sequence modeling with lower runtime overhead but may underperform on fine grained restoration tasks. To balance accuracy and efficiency, we train lightweight SSM blocks as feature-distilled surrogates of transformer blocks and use them to construct hybrid U-Net-style architectures. To automatically discover effective block combinations, we introduce Efficient Network Search (ENS), a multi-objective search strategy that selects task-specific hybrid configurations from pre-aligned components. ENS optimizes restoration quality while penalizing transformer usage, serving as a lightweight proxy for latency and enabling architecture discovery without repeated hardware profiling. On a Snapdragon 8 Elite CPU, the Restormer baseline requires 10119.52 ms for inference. In contrast, ENS-discovered hybrids significantly reduce runtime: ENS-Deblurring runs in 2973 ms (3.4x faster), ENS-Deraining in 5816 ms (1.74x faster), and ENS-Denoising in 8666 ms (1.17x faster), while maintaining competitive restoration quality.

Place recognition is essential for long-term autonomous navigation, enabling loop closure and consistent mapping. Although deep learning has improved performance, deploying such models on resource-constrained platforms remains challenging. This work explores efficient LiDAR-based place recognition for EdgeAI by leveraging Bird's Eye View representations to enable lightweight image-based networks. We benchmark representative architectures without aggregation heads using a unified descriptor scheme based on global pooling and linear projection, and evaluate performance under FP32, FP16, and INT8 quantization. Experiments reveal trade-offs between accuracy, robustness, and efficiency: FP16 matches FP32 with lower cost, while INT8 introduces architecture-dependent degradation. Overall, the presented results are a strong basis for future research on 'use-case'-aware quantisation of Neural Networks for Edge deployment.

Real-world fine manipulation, particularly in bimanual manipulation, typically requires low-latency control and stable visual localization, while collecting large-scale data is costly and limited demonstrations may lead to localization drift. Existing approaches make different trade-offs: action-chunking policies such as ACT enable low-latency execution and data efficiency but rely on dense visual features without explicit spatial consistency, generative methods such as Diffusion Policy improve expressiveness but can incur iterative sampling latency, vision-language-action and voxel-based methods enhance generalization and geometric grounding but require higher computational cost and system complexity. We introduce a multistage spatial attention module that extracts stable 2D attention points and jointly predicts future attention sequences with a temporal alignment loss. Built upon ACT with a pretrained ResNet visual prior, a multistage attention module extracts task-relevant 2D attention points as a local spatial modality for action prediction. To maintain consistent object tracking, we introduce a self-supervised objective that aligns predicted attention sequences with visual features from future frames, suppressing drift without keypoint annotations and improving stability of the vision-to-action mapping under limited data. Experiments on simulated and real-world fine manipulation tasks, conducted on the ALOHA bimanual platform, evaluate task success, attention drift, inference latency, and robustness to visual disturbances. Results indicate improvements in localization stability and task performance while maintaining low-latency inference under the tested conditions.

In wireless communications, recovering the optimal solution to the multiple-input multiple-output (MIMO) detection problem is NP-hard. Obtaining high-quality suboptimal solutions with a favorable performance-complexity trade-off is particularly challenging in under-determined systems with $N_t$ transmit antennas and $N_r < N_t$ receive antennas. Recent diffusion-based MIMO detectors have shown promise, but they require extensive sampling iterations at inference time, and their performance degrades in under-determined scenarios. We propose GD4, a graph-based discrete denoising diffusion method for MIMO detection. Unlike existing diffusion-based detectors that operate in a continuous relaxed space, GD4 performs denoising directly in the discrete symbol space and enables fast inference with one or a few denoising evaluations. Numerical results show that, under a similar inference-time compute budget, GD4 produces higher-quality suboptimal solutions than existing diffusion-based detectors and some widely used classical baseline including box-constrained Babai point and the $K$-best box-constrained randomized Klein-Babai point in both under-determined and overdetermined settings.

Reconstructing 3D scenes from sparse, unposed images remains challenging under real-world conditions with varying illumination and transient occlusions. Existing methods rely on scene-specific optimization using appearance embeddings or dynamic masks, which requires extensive per-scene training and fails under sparse views. Moreover, evaluations on limited scenes raise questions about generalization. We present GenWildSplat, a feed-forward framework for sparse-view outdoor reconstruction that requires no per-scene optimization. Given unposed internet images, GenWildSplat predicts depth, camera parameters, and 3D Gaussians in a canonical space using learned geometric priors. An appearance adapter modulates appearance for target lighting conditions, while semantic segmentation handles transient objects. Through curriculum learning on synthetic and real data, GenWildSplat generalizes across diverse illumination and occlusion patterns. Evaluations on PhotoTourism and MegaScenes benchmark demonstrate state-of-the-art feed-forward rendering quality, achieving real-time inference without test-time optimization

Large Language Models (LLMs) have become an integral part of many real-world workflows. However, LLMs consume a lot of energy, which becomes a large concern in the scale of the demand for these tools. As LLMs become integrated into different workflows, different applications have arisen to deal with the challenge of running inference for these tools. This raises another issue of choosing the runtime parameter values for these services in order to minimize the energy consumption. Oftentimes this requires deep knowledge of the application or traditional optimization methods that can take days to find optimal values. In this work, we created a human-in-the-loop flow with LLM-assisted runtime parameter optimization in order to solve this issue. With human-created, specific feedback prompting methods, chat-based LLMs can iteratively find energy-efficient inference parameters faster than traditional search methods. LLMs can also tailor their solutions to different hardware setups and easily take into account other system constraints. The enhanced prompt template was able to converge below the threshold at an average of 3.4 prompts compared to the baseline, which converged in an average of 5.2 prompts, and consistently achieved lower final energy per token. The enhanced prompt template also outperformed Sobol sampling in convergence speed.

We present Density Field State Space Models (DF-SSM), a framework for compressing SSMs to a 1-bit scaffold with int8 low-rank correction. Applied to Mamba-2 1.3B, we achieve a 278 MB model (9.7x smaller than the 2.7 GB FP16 teacher) that runs at 21.4x faster inference on GPU (batch=1, relative to the mamba-ssm reference implementation) while maintaining downstream task performance within 2-4 percentage points of BitMamba-2, a 1.58-bit model trained from scratch on 150B tokens. The distillation itself requires only 32M tokens and 6 hours on a single A100 GPU, though it presupposes a pretrained FP16 teacher. We develop an optimized inference pipeline combining cuBLAS INT8 tensor cores for the scaffold matmul, custom CUDA kernels for stateful SSM and convolution operations, and an AVX-512 CPU backend for efficient deployment on both GPU and CPU. Beyond compression, we investigate the internal knowledge organization of the resulting model, discovering three distinct processing phases: intent classification (layers 0-3, operating in an abstract space with no vocabulary alignment), knowledge retrieval (layers 25-35, where factual associations localize to a 5-layer window), and output formatting (layers 36-47, where category structure dissolves). Through systematic analysis of 445 factual prompts across 19 categories, we find that early-layer classification is syntactic (driven by template structure) rather than semantic, and that the model exhibits well-organized knowledge representations despite weak factual recall--suggesting that representational structure may precede factual strength.

Real-time automatic speech recognition (ASR) systems face a fundamental trade-off between transcription accuracy and computational efficiency, particularly when deploying large-scale transformer models like Whisper. Existing streaming approaches either sacrifice accuracy through aggressive chunking or incur prohibitive memory costs through unbounded context accumulation. We present WhisperPipe, a novel streaming architecture that achieves bounded memory consumption while maintaining transcription quality through three key innovations a hybrid Voice Activity Detection (VAD) pipeline combining Silero VAD with energy-based filtering to reduce false activations by 34%, a dynamic buffering mechanism with overlapping context windows that prevents information loss at segment boundaries, and an adaptive processing strategy that balances latency and accuracy based on speech characteristics. Evaluated on 2.5 hours of diverse audio data, WhisperPipe demonstrates a median end-to-end latency of 89ms (90th percentile: 142ms) while consuming 48% less peak GPU memory and 80.9% lower average GPU utilization compared to baseline Whisper implementations. The system maintains stable memory usage over extended sessions, with zero growth rate across 150-minute continuous operation. Comparative analysis against related work shows that WhisperPipe achieves competitive accuracy (WER within 2% of offline Whisper) while operating at 3-5x lower latency than existing streaming solutions. The architecture's modular design enables deployment across resource-constrained environments, from edge devices to cloud infrastructure. Our results demonstrate that careful architectural design can reconcile the competing demands of real-time responsiveness and model sophistication in production ASR systems.

Unlike chatbots, physical AI must act while the world keeps evolving. Therefore, the inter-chunk pause of synchronous executors are fatal for dynamic tasks regardless of how fast the inference is. Asynchronous execution -- thinking while acting -- is therefore a structural requirement, and real-time chunking (RTC) makes it viable by recasting chunk transitions as inpainting: freezing committed actions and consistently generating the remainder. However, RTC with flow-matching policy is structurally suboptimal: its inpainting comes from inference-time corrections rather than the base policy, yielding little pre-training benefit, specific fine-tuning, heuristic guidance, and extra computation that inflates the latency. In this work, we observe that discrete diffusion policies, which generate actions by iteratively unmasking, are natural asynchronous executors that resolve all limitations at once: they are fine-tuning free since inpainting is their native operation, while early stopping further provides adaptive guidance and reduces inference cost. We propose DiscreteRTC, which replaces external corrections with native unmasking, and show on dynamic simulated benchmarks and real-world dynamic manipulation tasks that it achieves higher success rates than continuous RTC and other baselines. In summary, DiscreteRTC is simpler to implement with 0 lines of additional code to enable async inpainting, faster at inference with only ~0.7 computation compared with generating actions from scratch, and better at execution with 65% higher success rate in real-world hockey defend task compared with flow-matching RTC, and 30% higher compared with training-time flow-matching RTC. More visualizations are on https://outsider86.github.io/DiscreteRTCSite/.

Vision-Language-Action (VLA) models are promising for generalist robot control, but on-robot deployment is bottlenecked by real-time inference under tight cost and energy budgets. Most prior evaluations rely on desktop-grade GPUs, obscuring the trade-offs and opportunities offered by heterogeneous edge accelerators (GPUs/XPUs/NPUs). We present a systematic analysis for low-cost VLA deployment via model-hardware co-characterization. First, we build a cross-accelerator leaderboard and evaluate model-hardware pairs under CET (Cost, Energy, Time), showing that right-sized edge devices can be more cost-/energy-efficient than flagship GPUs while meeting control-rate constraints. Second, using in-depth profiling, we uncover a consistent two-phase inference pattern: a compute-bound VLM backbone followed by a memory-bound Action Expert, which induces phase-dependent underutilization and hardware inefficiency. Finally, guided by these insights, we propose DP-Cache and V-AEFusion to reduce diffusion redundancy and enable asynchronous pipeline parallelism, achieving up to 2.9x speedup on GPUs and 6x on edge NPUs with only marginal success degradation. The example leaderboard website is available at: https://vla-leaderboard-01.vercel.app/.

Efficiently solving the Job Shop Scheduling Problem in real-world industrial applications requires policies that are both computationally lean and topologically robust. While Reinforcement Learning has shown potential in automating dispatching rules, existing models often struggle with a scalability bottleneck caused by quadratic graph complexity or the architectural overhead of heterogeneous layers. We introduce a unified graph framework that employs feature-based homogenization to project distinct node roles into a shared latent space. This allows a standard homogeneous Graph Isomorphism Network to capture complex resource contention with linear complexity, ensuring low-latency inference for large-scale industrial applications. Our empirical results demonstrate that our framework achieves state-of-the-art performance while exhibiting consistent zero-shot generalization. We identify the job-to-machine ratio as the primary driver of policy effectiveness, rather than absolute problem size. Based on this, we propose a hypothesis of structural saturation, demonstrating that policies trained on critically congested instances ($\mathcal{J} \approx \mathcal{M}$) learn scale-invariant resolution strategies. Agents trained at this saturation point internalize invariant conflict-resolution logic, allowing them to treat massive rectangular instances as a sequential concatenation of saturated sub-problems. This approach eliminates the need for expensive scale-specific retraining and prevents overfitting to statistical shortcuts, providing a robust and efficient pathway for deploying RL solutions in dynamic production environments.

Existing attention accelerators often trade exact softmax semantics, depend on fused Tensor Core kernels, or incur sequential depth that limits FP32 throughput on long sequences. We present \textbf{ELSA}, an algorithmic reformulation of online softmax attention that (i)~preserves exact softmax semantics in real arithmetic with a \emph{provable} $\mathcal{O}(u\log n)$ FP32 relative error bound; (ii)~casts the online softmax update as a prefix scan over an associative monoid $(m,S,W)$, yielding $O(n)$ extra memory and $O(\log n)$ parallel depth; and (iii)~is Tensor-Core independent, implemented in Triton and CUDA C++, and deployable as a \emph{drop-in replacement} requiring no retraining or weight modification. Unlike FlashAttention-2/3, which rely on HMMA/GMMA Tensor Core instructions and provide no compatible FP32 path, ELSA operates identically on A100s and resource-constrained edge devices such as Jetson TX2 -- making it the only hardware-agnostic exact-attention kernel that reduces parallel depth to $O(\log n)$ at full precision. On A100 FP32 benchmarks (1K--16K tokens), ELSA delivers $1.3$--$3.5\times$ speedup over memory-efficient SDPA and $1.97$--$2.27\times$ on BERT; on Jetson TX2, ELSA achieves $1.5$--$1.6\times$ over Math (64--900 tokens), with $17.8$--$20.2\%$ throughput gains under LLaMA-13B offloading at $\ge$32K. In FP16, ELSA approaches hardware-fused baselines at long sequences while retaining full FP32 capability, offering a unified kernel for high-precision inference across platforms. Our code and implementation are available at https://github.com/ming053l/ELSA.

This paper presents a complete, end-to-end on-device vision machine learning pipeline, comprising data acquisition, two-layer CNN training with Adam optimization, and real-time inference, executing entirely on a microcontroller-class device costing $15-40 USD. Unlike cloud-based workflows that require external infrastructure and conceal the computational pipeline from the practitioner, this system implements every step of the core ML lifecycle in approximately 1,750 lines of readable C++ that compiles in under one minute using the Arduino IDE, with no external ML dependencies. Running on the Seeed Studio ESP32-S3 XIAO ML Kit (8 MB PSRAM), the firmware achieves three-class 64x64 image classification in approximately 9 minutes per training run, with real-time inference at 6.3 FPS. Key contributions include: correct batch-level gradient accumulation; pre-computed resize lookup tables for inference; dual-format weight export for SD-free baked-in deployment; a three-tier weight priority system (SD binary > baked-in header > He-initialization) resolved automatically at boot; a single-constant network reconfiguration interface; and PSRAM-aware memory management suited to microcontroller constraints. All source code and reference datasets are released under the MIT License at https://github.com/webmcu-ai/on-device-vision-ai