Computational decarbonization aims to reduce carbon emissions in computing and societal systems such as data centers, transportation, and built environments. This requires accurate, fine-grained carbon intensity forecasts, yet existing tools have several key limitations: (i) they require grid-specific electricity mix data, restricting use where such information is unavailable; (ii) they depend on separate grid-specific models that make it challenging to provide global coverage; and (iii) they provide forecasts without uncertainty estimates, limiting reliability for downstream carbon-aware applications. In this paper, we present CarbonX, an open-source tool that leverages Time Series Foundation Models (TSFMs) for a range of decarbonization tasks. CarbonX utilizes the versatility of TSFMs to provide strong performance across multiple tasks, such as carbon intensity forecasting and imputation, and across diverse grids. Using only historical carbon intensity data and a single general model, our tool achieves a zero-shot forecasting Mean Absolute Percentage Error (MAPE) of 15.82% across 214 grids worldwide. Across 13 benchmark grids, CarbonX performance is comparable with the current state-of-the-art, with an average MAPE of 9.59% and tail forecasting MAPE of 16.54%, while also providing prediction intervals with 95% coverage. CarbonX can provide forecasts for up to 21 days with minimal accuracy degradation. Further, when fully fine-tuned, CarbonX outperforms the statistical baselines by 1.2--3.9X on the imputation task. Overall, these results demonstrate that CarbonX can be used easily on any grid with limited data and still deliver strong performance, making it a practical tool for global-scale decarbonization.

Foundation models (FMs) are increasingly used as backbones for downstream tasks across language, vision, time-series, and multimodal applications. Yet existing model-serving systems deploy each customized task as an independent model instance, thereby replicating heavyweight backbones, wasting accelerator memory, and losing opportunities to amortize batching and loading costs. This paper presents FMplex, a serving system that treats FM backbones as a virtualization substrate for deployment sharing. FMplex presents each task with a virtual foundation model (vFM), a logically private FM instance backed by a shared physical FM. This abstraction lets independently customized tasks share a backbone while preserving task-specific extensions, independent lifecycles, and task-level isolation. In addition, we propose a batch-aware fair-queueing scheduler that combines weighted task-level sharing with inter- and intra-task batching across colocated tasks. We implement a FMplex-based serving stack spanning task construction, sharing-aware deployment, and runtime execution. Across 7 FM backbones (16 variants) and 92 downstream tasks, FMplex reduces latency by up to 80

Large Vision–Language Models (LVLMs) use their vision encoders to translate images into representations for downstream reasoning, but the encoders often underperform in domain-specific visual tasks such as medical image diagnosis or fine-grained classification, where representation errors can cascade through the language model, leading to incorrect responses.Existing adaptation methods modify the continuous feature interface between encoder and language model through projector tuning or other parameter-efficient updates, which still couples the two components and requires re-alignment whenever the encoder changes.We introduce CRAFT (Codebook RegulAted Fine-Tuning), a lightweight method that fine-tunes the encoder using a discrete codebook that anchors visual representations to a stable token space, achieving domain adaptation without modifying other parts of the model.This decoupled design allows the adapted encoder to seamlessly boost the performance of LVLMs with different language architectures, as long as they share the same codebook. Empirically, CRAFT achieves an average gain of 14.98\% across 10 domain-specific benchmarks such as VQARAD and PlantVillage, while preserving the LLM’s linguistic capabilities and outperforming peer methods that operate on continuous tokens.

Deep learning models are increasingly central to autonomous vehicle (AV) pipelines, yet their integration has traditionally followed a monolithic design where perception, planning, and control execute on a single onboard computer. This design overlooks the emerging paradigm of cooperative autonomy, where vehicles interact with roadside units (RSUs), edge servers, and cloud-hosted intelligence through vehicle-to-everything (V2X) connectivity. Cooperative perception and control improve safety and efficiency, but also introduce systems-level challenges: network latency, compute heterogeneity, and multi-tenant contention, all critically affect real-time decision-making. These challenges are further amplified by the increasing reliance on large foundation models, whose scale necessitates cloud deployment. We present CADET (Cooperative Autonomy through Distributed Experimentation Toolkit), a modular platform for systematic and reproducible evaluation of distributed cooperative autonomy systems under realistic deployment conditions. CADET decouples the AV stack into composable modules that can be flexibly deployed across vehicles, infrastructure, and edge/cloud tiers. The framework integrates state-of-the-art models, incorporates trace-driven network and workload emulation, and provides synchronized model-, system-, and task-level instrumentation. Through V2V and V2I experiments, we show that distributed deployment choices fundamentally shape safety, with V2V intent packets outperforming cloud-based perception and RSU-assisted perception sustaining safety until overloaded by concurrent requests. Although designed for AV pipelines, CADET also supports dataset-driven experimentation, enabling systems and ML researchers to benchmark distributed inference workloads independently of full vehicle simulation. CADET is open source, with code and demo available at https://nesl.github.io/cadet-web.

Large language models (LLMs) and time-series language models (TSLMs) are increasingly applied to time-series question answering (TSQA). Unlike text-only QA, TSQA requires models to ground answers in temporal signals whose patterns may occur at different scales, specific time locations, or across separated intervals. However, existing benchmarks are typically organized by task types or high-level reasoning categories, making it difficult to diagnose the underlying signal-level capabilities driving model performance. We introduce TS-Skill, a controlled benchmark for evaluating three composable analytical skills in TSQA: temporal scale selection (SK1), temporal localization (SK2), and cross-interval integration (SK3). TS-Skill provides timestamp-aware questions, broad domain coverage, and human-validated QA quality. To construct the benchmark at scale, we develop SKEvol, a skill-guided agentic framework that combines domain-aware time-series seed generation, skill-controlled question generation, metadata- and code-assisted answer construction, multi-phase signal-grounded verification, and human-in-the-loop curation. Experiments on ten state-of-the-art LLMs and TSLMs reveal substantial and uneven capability gaps across SK1-SK3. In particular, SK3 remains consistently challenging for non-agent models, whereas tool-augmented agents show a selective advantage on standalone SK3. These findings demonstrate that skill-level evaluation can uncover temporal reasoning failures that are obscured by aggregate TSQA scores.

Deploying neural networks on low-power microcontrollers (MCUs) requires selecting model architectures under tight memory, latency, and energy constraints. Existing workflows often simplify this process along one or more axes: static proxy costs such as FLOPs or parameters, treating one MCU as representative, and continuous-inference tests instead of deployed sensing schedules. These assumptions can mis-rank Pareto-front candidates, miss infeasible deployments, and obscure schedule-dependent energy. We present CREST (Cross-platform Runtime Evaluation and Search Tool), a deployment-realistic hardware-in-the-loop (HIL) neural architecture search (NAS) framework for MCU sensing systems. CREST keeps the optimizer, HIL measurement boundary, logging, and replay workflow fixed while exposing workload, model family, target backend, schedule, quantization, and scoring policy as configurable axes. This makes deployment effects experimentally separable within one reusable workflow. We evaluate CREST on inertial odometry and audio classification across three Arm Cortex-M targets. For inertial odometry, measured-energy HIL search reduces median per-inference energy by 41.7% versus FLOPs-based selection and 40.8% versus memory-traffic-based selection at similar error. FLOPs-based selection also chooses infeasible deployments on memory-constrained targets. On the STM32 N657 target, continuous-inference and duty-cycled searches produce different Pareto frontiers. For audio classification, the same application-level policy selects different DS-CNN architectures on different boards, and cross-board replay changes deployment cost substantially. Overall, CREST shows that deployment-realistic MCU NAS must jointly optimize model architecture, target platform, runtime schedule, and deployment policy rather than relying only on static proxy costs or continuous-inference measurements.

As edge AI deployments scale to billions of devices running always-on, real-time compound AI pipelines, they represent a massive and largely unmanaged source of energy consumption and carbon emissions. To reduce carbon emissions while maximizing Quality-of-Service (QoS), this paper proposes FM-CAC, a proactive carbon-aware control framework that leverages a battery as an active temporal buffer. By decoupling energy acquisition from energy consumption, FM-CAC can maximize the use of low-carbon energy, substantially reducing carbon emissions. At each control step, FM-CAC jointly optimizes the software pipeline variant, the hardware operating point, and the battery charging and discharging actions. To support this decision process, FM-CAC leverages edge-friendly Time-Series Foundation Models (TSFMs) for zero-shot carbon forecasting and integrates these forecasts into a dynamic programming solver with deferred cost attribution to prevent myopic battery depletion. Results show that FM-CAC reduces carbon emissions by up to 65.6

Evolutionary Algorithms (EA) have recently been explored as a mechanism for improving large language model (LLM) code generation, primarily in software engineering, programming competitions, and combinatorial optimization. However, their efficacy in Sensor Data Processing (SDP) remains largely unexplored. SDP tasks exhibit unique characteristics, including signal assumptions, numerical sensitivity, and distinct problem abstraction, that significantly differ from conventional coding benchmarks. It is therefore unclear whether EA can reliably improve performance in this domain.In this work, we adopt the EA-based program search and apply it to coding-based SDP. We systematically evaluate its impact across diverse SDP tasks, such as spectral filtering, imputation, localization, and tracking. Our study compares standard coding baselines and EA-enhanced program search, analyzing not only accuracy but also robustness, consistency across runs, and failure modes.Our results show that EA generally improves performance and produces more reliable solutions for many sensor tasks. At the same time, the improvements are not universal due to cross-run variations in implementations. In addition, through failure case studies, we show that many errors stem from incorrect signal assumptions rather than from failures of the search process itself. Overall, this work provides an empirical evaluation of applying EA to SDP, highlighting its strengths and limitations for more reliable sensor processing agents in the future.

Effective sensor data processing is critical for cyber-physical and Internet of Things (IoT) systems, but often requires specialized expertise. While large language models (LLMs) show promise as autonomous copilots for sensor processing, their capabilities remain underexplored. We introduce SensorBench, the first comprehensive benchmark for evaluating LLMs across diverse real-world sensor datasets and tasks. SensorBench evaluates three paradigms for leveraging LLMs in sensing tasks: tool-augmented coding (TAC), standalone coding (SAC), and direct answer (DA). We evaluate 8 leading LLM variants, including 2 large reasoning models (LRMs) and 2 domain-specific LLMs, providing a structured reference for absolute performance, latency, and resource requirements. Our analysis reveals that: 1) TAC significantly outperforms SAC and DA; 2) LLMs excel at simple tasks but consistently underperform domain experts on compositional tasks requiring parameter tuning and multistep reasoning; and 3) the reasoning mechanism introduced in LRMs does not yield substantial performance gains. To improve the performance, we explore four prompting strategies and fine-tuning approaches (using our newly released sensor-processing corpus). The results show that self-verification prompting proves most effective, outperforming other methods simultaneously in 48% of tasks, while fine-tuning yields marginal gains. Our analysis suggests that more sophisticated interaction frameworks, such as signal-level self-verification, may bridge the gap to human expert-level performance. This benchmark provides a foundation for evaluating and improving LLMs in sensing applications https://github.com/nesl/LLM_sensor_processing

Radio-frequency (RF) data synthesis predicts the received signal given transmitter and receiver positions, and is essential for wireless applications. Recent 3D Gaussian Splatting (3DGS)-based methods achieve efficient synthesis at any transmitter but only for a fixed receiver. Therefore, supporting N receivers in one scene requires N independent models and precludes prediction at unseen receivers. We present RxGS, which achieves receiver-generalizable synthesis within a single unified model. Our key insight is that scene geometry is receiver-independent while directional radiance is not: a first stage learns shared 3D Gaussian geometry, and a second stage freezes it and learns directional radiance conditioned on receiver position. A global conditioning branch captures shared receiver-dependent effects across the scene, while a local branch models per-scatterer variations from the receiver's geometry and occlusion. A multi-receiver CUDA rasterizer further batches rendering across all N receivers. Evaluated across various RF datasets, RxGS matches or improves over per-receiver baselines with a single shared model and generalizes to receivers unseen during training within the scene, cutting training cost by up to 45×, inference cost by 7.6×, and storage by N×.

Building energy management (BEM) tasks require processing and learning from a variety of time-series data. Existing solutions rely on bespoke task- and data-specific models to perform these tasks, limiting their broader applicability. Inspired by the transformative success of Large Language Models (LLMs), Time-Series Foundation Models (TSFMs), trained on diverse datasets, have the potential to change this. Were TSFMs to achieve a level of generalizability across tasks and contexts akin to LLMs, they could fundamentally address the scalability challenges pervasive in BEM. To understand where they stand today, we evaluate TSFMs across four dimensions: (1) generalizability in zero-shot univariate forecasting, (2) forecasting with covariates for thermal behavior modeling, (3) zero-shot representation learning for classification tasks, and (4) robustness to performance metrics and varying operational conditions. Our results reveal that TSFMs exhibit limited generalizability, performing only marginally better than statistical models on unseen datasets and modalities for univariate forecasting. Similarly, inclusion of covariates in TSFMs does not yield performance improvements, and their performance remains inferior to conventional models that utilize covariates. While TSFMs generate effective zero-shot representations for downstream classification tasks, they may remain inferior to statistical models in forecasting when statistical models perform test-time fitting. Moreover, TSFMs forecasting performance is sensitive to evaluation metrics, and they struggle in more complex building environments compared to statistical models. These findings underscore the need for targeted advancements in TSFM design, particularly their handling of covariates and incorporating context and temporal dynamics into prediction mechanisms, to develop more adaptable and scalable solutions for BEM.

Existing Wi-Fi sensing systems rely on injecting high-rate probing packets to extract channel state information (CSI), leading to communication degradation and poor deployability. Although Integrated Sensing and Communication (ISAC) is a promising direction, existing solutions still rely on auxiliary packet injection because they exploit only CSI from data frames. We present UniFi, the first Wi-Fi-based ISAC framework that fully eliminates intrusive packet injection by directly exploiting irregularly sampled CSI from diverse communication packets across multiple frequency bands. UniFi integrates a CSI sanitization pipeline to harmonize heterogeneous packets and remove burst-induced redundancy, together with a time-aware attention model that learns directly from non-uniform CSI sequences without resampling. We further introduce CommCSI-HAR, the first dataset with irregularly sampled CSI from real-world dual-band communication traffic. Extensive evaluations on this dataset and four public benchmarks show that UniFi achieves state-of-the-art accuracy with a compact model size, while fully preserving communication throughput.

Accurate observation of dynamic environments traditionally relies on synthesizing raw, signal-level information from multiple distributed sensors. This work investigates an alternative approach: performing geospatial inference using only encrypted packet-level information, without access to the raw sensory data. We further explore how this indirect information can be fused with directly available sensory data to extend overall inference capabilities. We introduce GraySense, a learning-based framework that performs geospatial object tracking by analyzing encrypted wireless video transmission traffic, such as packet sizes, from cameras with inaccessible streams. GraySense leverages the inherent relationship between scene dynamics and transmitted packet sizes to infer object motion. The framework consists of two stages: (1) a Packet Grouping module that identifies frame boundaries and estimates frame sizes from encrypted network traffic, and (2) a Tracker module, based on a Transformer encoder with a recurrent state, which fuses indirect packet-based inputs with optional direct camera-based inputs to estimate the object's position. Extensive experiments with realistic videos from the CARLA simulator and emulated networks under varying conditions show that GraySense achieves 2.33 meters tracking error (Euclidean distance) without raw signal access, within the dimensions of tracked objects (4.61m x 1.93m). To our knowledge, this capability has not been previously demonstrated, expanding the use of latent signals for sensing.

Multimodal deep neural networks deployed in realistic environments must contend with runtime variations: changes in modality quality, overall input complexity, and available platform resources. Current networks struggle with such fluctuations – adaptive networks cannot adhere to a strict compute budget, controller-based networks neglect to consider input complexity, and statically provisioned networks fail at all the above. Consequently, they do not extract maximum utility from the expended computational resources. We present SWAN (Sample and World-Aware Multimodal Network), the first adaptive multimodal network that accomplishes all three goals. SWAN employs a quality-aware controller to assign resources among modalities according to a variable user-specified maximum budget. Within this budget, an adaptive gating module further optimizes efficiency by scaling layer utilization according to sample complexity. For further gains, SWAN also employs a token dropping module that masks semantically irrelevant multimodal features before performing detections. We evaluate SWAN in the domain of autonomous driving with complex multi-object 3D detection, reducing FLOPs by up to 49

Multimodal deep learning systems are deployed in dynamic scenarios due to the robustness afforded by multiple sensing modalities. Nevertheless, they struggle with varying compute resource availability (due to multi-tenancy, device heterogeneity, etc.) and fluctuating quality of inputs (from sensor feed corruption, environmental noise, etc.). Statically provisioned multimodal systems cannot adapt when compute resources change over time, while existing dynamic networks struggle with strict compute budgets. Additionally, both systems often neglect the impact of variations in modality quality. Consequently, modalities suffering substantial corruption may needlessly consume resources better allocated towards other modalities. We propose ADMN, a layer-wise Adaptive Depth Multimodal Network capable of tackling both challenges - it adjusts the total number of active layers across all modalities to meet compute resource constraints, and continually reallocates layers across input modalities according to their modality quality. Our evaluations showcase ADMN can match the accuracy of state-of-the-art networks while reducing up to 75% of their floating-point operations.

Modern urban spaces are equipped with an increasingly diverse set of sensors, all producing an abundance of multimodal data. Such multimodal data can be used to identify and reason about important incidents occurring in urban landscapes, such as major emergencies, cultural and social events, as well as natural disasters. However, such data may be fragmented over several sources and difficult to integrate due to the reliance on human-driven reasoning for identifying relationships between the multimodal data corresponding to an incident, as well as understanding the different components which define an incident. Such relationships and components are critical to identifying the causes of such incidents, as well as producing forecasting the scale and intensity of future incidents as they begin to develop. In this work, we create SIGMUS, a system for Semantic Integration for Knowledge Graphs in Multimodal Urban Spaces. SIGMUS uses Large Language Models (LLMs) to produce the necessary world knowledge for identifying relationships between incidents occurring in urban spaces and data from different modalities, allowing us to organize evidence and observations relevant to an incident without relying and human-encoded rules for relating multimodal sensory data with incidents. This organized knowledge is represented as a knowledge graph, organizing incidents, observations, and much more. We find that our system is able to produce reasonable connections between 5 different data sources (new article text, CCTV images, air quality, weather, and traffic measurements) and relevant incidents occurring at the same time and location.

Complex events (CEs) play a crucial role in CPS-IoT applications, enabling high-level decision-making in domains such as smart monitoring and autonomous systems. However, most existing models focus on short-span perception tasks, lacking the long-term reasoning required for CE detection. CEs consist of sequences of short-time atomic events (AEs) governed by spatiotemporal dependencies. Detecting them is difficult due to long, noisy sensor data and the challenge of filtering out irrelevant AEs while capturing meaningful patterns. This work explores CE detection as a case study for CPS-IoT foundation models capable of long-term reasoning. We evaluate three approaches: (1) leveraging large language models (LLMs), (2) employing various neural architectures that learn CE rules from data, and (3) adopting a neurosymbolic approach that integrates neural models with symbolic engines embedding human knowledge. Our results show that the state-space model, Mamba, which belongs to the second category, outperforms all methods in accuracy and generalization to longer, unseen sensor traces. These findings suggest that state-space models could be a strong backbone for CPS-IoT foundation models for long-span reasoning tasks.

Today's Internet of Things (IoT) has evolved from simple sensing and actuation devices to those with embedded processing and intelligent services, enabling rich collaborations between users and their devices. However, enabling such collaboration becomes challenging when transient devices need to interact with host devices in temporarily visited environments. In such cases, fine-grained access control policies are necessary to ensure secure interactions; however, manually implementing them is often impractical for non-expert users. Moreover, at run-time, the system must automatically configure the devices and enforce such fine-grained access control rules. Additionally, the system must address the heterogeneity of devices. In this paper, we present CollabIoT, a system that enables secure and seamless device collaboration in transient IoT environments. CollabIoT employs a Large language Model (LLM)-driven approach to convert users' highlevel intents to fine-grained access control policies. To support secure and seamless device collaboration, CollabIoT adopts capability-based access control for authorization and uses lightweight proxies for policy enforcement, providing hardware-independent abstractions. We implement a prototype of CollabIoT's policy generation and auto configuration pipelines and evaluate its efficacy on an IoT testbed and in large-scale emulated environments. We show that our LLM-based policy generation pipeline is able to generate functional and correct policies with 100% accuracy. At runtime, our evaluation shows that our system configures new devices in similar to 150 ms, and our proxy-based data plane incurs network overheads of up to 2 ms and access control overheads up to 0.3 ms.

Practitioners deploying time series forecasting models face a dilemma: exhaustively validating dozens of models is computationally prohibitive, yet choosing the wrong model risks poor performance. We show that spectral predictability~$Ω$ -- a simple signal processing metric -- systematically stratifies model family performance, enabling fast model selection. We conduct controlled experiments in four different domains, then further expand our analysis to 51 models and 28 datasets from the GIFT-Eval benchmark. We find that large time series foundation models (TSFMs) systematically outperform lightweight task-trained baselines when $Ω$ is high, while their advantage vanishes as $Ω$ drops. Computing $Ω$ takes seconds per dataset, enabling practitioners to quickly assess whether their data suits TSFM approaches or whether simpler, cheaper models suffice. We demonstrate that $Ω$ stratifies model performance predictably, offering a practical first-pass filter that reduces validation costs while highlighting the need for models that excel on genuinely difficult (low-$Ω$) problems rather than merely optimizing easy ones.

The increasing deployment of deep neural networks (DNNs) in cyber-physical systems (CPS) enhances perception fidelity, but imposes substantial computational demands on execution platforms, posing challenges to real-time control deadlines. Traditional distributed CPS architectures typically favor on-device inference to avoid network variability and contention-induced delays on remote platforms. However, this design choice places significant energy and computational demands on the local hardware. In this work, we revisit the assumption that cloud-based inference is intrinsically unsuitable for latency-sensitive control tasks. We demonstrate that, when provisioned with high-throughput compute resources, cloud platforms can effectively amortize network and queueing delays, enabling them to match or surpass on-device performance for real-time decision-making. Specifically, we develop a formal analytical model that characterizes distributed inference latency as a function of the sensing frequency, platform throughput, network delay, and task-specific safety constraints. We instantiate this model in the context of emergency braking for autonomous driving and validate it through extensive simulations using real-time vehicular dynamics. Our empirical results identify concrete conditions under which cloud-based inference adheres to safety margins more reliably than its on-device counterpart. These findings challenge prevailing design strategies and suggest that the cloud is not merely a feasible option, but often the preferred inference location for distributed CPS architectures. In this light, the cloud is not as distant as traditionally perceived; in fact, it is closer than it appears.