Machine Unlearning aims to remove the influence of specific data or concepts from trained models while preserving overall performance, a capability increasingly required by data protection regulations and responsible AI practices. Despite recent progress, unlearning in text-to-image diffusion models remains challenging due to high computational costs and the difficulty of balancing effective forgetting with retention of unrelated concepts. We introduce Self-distillation for PARameter Efficient Removal (SPARE), a two-stage unlearning method for image generation that combines parameter localization with self-distillation. SPARE first identifies parameters most responsible for generation of the unwanted concepts using gradient-based saliency and constrains updates through sparse low rank adapters, ensuring lightweight, localized modifications. In a second stage, SPARE applies a self-distillation objective that overwrites the unwanted concept with a user-defined surrogate while preserving behavior for other concepts. In addition we proposed a timestep sampling scheme for diffusion models to target only the crucial timesteps for a given concept leading to efficient unlearning. SPARE sur
This paper presents a Markov-chain-based method for the early-phase analysis and design of hybrid spare-management architectures for large-scale satellite constellations.} The hybrid strategy combines two channels: an indirect path that stages spares in parking orbits via heavy launch for later transfer to constellation planes, and a direct path that delivers spares to in-plane orbits using small launch vehicles. {To assess the long-run viability of such concepts of operations, satellite failure and replenishment processes are modeled as a Markov chain:} the indirect channel follows a periodic-review reorder-point/order-quantity policy, while the direct channel uses a standard reorder-point/order-quantity policy. These coupled chains yield a periodic steady state over the right ascension of the ascending node cycle via fixed-point iteration, and the stationary distributions provide rigorous cost and resilience metrics. By directly modeling the stochastic, multi-echelon dynamics governed by orbital mechanics, our framework avoids the aggregation assumptions of prior works and remains valid across a wider operating domain. We also introduce an approximate analysis that preserves dela
In large-scale LLM pre-training systems with 100k+ GPUs, failures become the norm rather than the exception, and restart costs can dominate wall-clock training time. However, existing fault-tolerance mechanisms are largely unprepared for this restart-dominant regime. To address this challenge, we propose SPARe - Stacked Parallelism with Adaptive Reordering - a fault-tolerance framework that masks node failures during gradient synchronization by stacking redundant data shards across parallelism groups and adaptively reordering execution. SPARe achieves availability comparable to traditional replication while maintaining near-constant computation overhead of only 2~3x, even under high redundancy where traditional replication would require linearly inflating overhead. We derive closed-form expressions for endurable failure count and computation overhead, validate them via SimGrid-based discrete-event simulation, and jointly optimize redundancy and checkpointing to minimize time-to-train. At extreme scale with up to 600k GPUs, SPARe reduces time-to-train by 40~50% compared to traditional replication.
Process or step-wise supervision has played a crucial role in advancing complex multi-step reasoning capabilities of Large Language Models (LLMs). However, efficient, high-quality automated process annotation remains a significant challenge. To address this, we introduce Single-Pass Annotation with Reference-Guided Evaluation (SPARE), a novel structured framework that enables efficient per-step annotation by jointly aligning solution steps to reference solutions and determine its accuracy with explicit reasoning in single generation. We demonstrate SPARE's effectiveness across four diverse datasets spanning mathematical reasoning (GSM8K, MATH), multi-hop question answering (MuSiQue-Ans), and spatial reasoning (SpaRP), showing consistent improvements in two applications: (1) training Process Reward Models (PRMs) for ranking and aggregating multiple generations, and (2) fine-tuning models via offline reinforcement learning for greedy decoding. On ProcessBench, SPARE demonstrates data-efficient out-of-distribution generalization, using only $\sim$16% of training samples compared to human-labeled and other synthetically trained baselines. Additionally, it achieves competitive performan
This paper presents a Markov-chain-based method for the early-phase analysis and design of spare-management architectures for large-scale satellite constellations. To assess the long-run viability of such concepts of operations, satellite failure and replenishment processes are modeled as Markov chains and analyzed through their stationary solution. We reinvestigate an indirect spare strategy, modeled as a multi-echelon periodic-review reorder-point/order-quantity policy, in which spares are first delivered to parking orbits and then transferred to constellation planes. The stock levels in constellation and parking orbits are each modeled as independent Markov chains, and a fixed-point iteration yields a consistent joint stationary solution that describes the strategy's average behavior. This approach accurately captures the stochastic interplay within a multi-echelon model driven by orbital mechanics, avoiding the aggregation assumptions of prior works and remaining valid across a wider operating domain. Building on this fast, accurate analysis, we formulate an optimization problem and solve it via a genetic algorithm. Finally, we demonstrate the practical value of both the analys
This paper introduces the analysis and design method of an optimal spare management policy using Markov chain for a large-scale satellite constellation. We propose an analysis methodology of spare strategy using a multi-echelon $(r,q)$ inventory control model with Markov chain, and review two different spare strategies: direct resupply, which inserts spares directly into the constellation orbit using launch vehicles; and indirect resupply, which places spares into parking orbits before transferring them to the constellation orbit. Furthermore, we propose an optimization formulation utilizing the results of the proposed analysis method, and an optimal solution is found using a genetic algorithm.
Loss of voluntary foot movement after spinal cord injury (SCI) can significantly limit independent mobility and quality of life. To improve motor output after injury, functional electrical stimulation (FES) is used to deliver stimulation pulses through the skin to affected muscles. While commercial FES systems typically use motion-based triggers, prior research shows that spared movement intent can be decoded after SCI using surface electromyography (EMG). Our aim is to assess how well spared neural signals of the lower limb after SCI can be decoded and used to control electrical stimulation for restoring foot movement. We developed a wearable machine learning-powered neuroprosthetic that records EMG from the affected lower limb using a 32-channel electrode bracelet and enables closed-loop control of a FES device for foot movement restoration. Five participants with SCI used the predicted control signal to follow trajectories on a screen with their foot and achieve distinct motor activation patterns for foot flexion, extension, and inversion or eversion. Three of these participants also achieved 2 proportional activation levels during foot flexion/extension with more than 70% accur
Maintenance organizations in manufacturing try to avoid downtime and unnecessary purchasing by reusing existing assets, but the main obstacle is not a lack of parts but a lack of actionable visibility across sites and partners. Inventories are distributed, described with inconsistent naming conventions, and contain duplicates and partially specified references, so the right part often exists somewhere but remains effectively undiscoverable. The paper proposes PhRAG, a hybrid Retrieval-Augmented Generation for pooling this fragmented landscape into a Virtual Stock Pool (VSPool) that can be structured and searched as a single resource. Heterogeneous spare part descriptions are structured via Named Entity Recognition (NER) into a shared virtual pool dataset and indexed to support robust retrieval even when users express needs in natural language rather than exact technical specifications. The proposed modular pipeline leverages the multitasking nature of generative language models to cover two dimensions that make industrial parts pooling challenging: ($\boldsymbol{i}$) unstructured technical specifications from diverse data sources (e.g. new partners, catalogs, marketplace listings)
Ultra-reliable low-latency vehicular communications (URLLC) require sufficient physical-layer (PHY) compute headroom at the network edge, where roadside units (RSUs) and compact next-generation base stations (gNBs) must meet strict timing constraints while co-hosting higher-layer services. In 5G New Radio (5G NR), low-density parity-check code (LDPC) decoding is a latency-sensitive iterative PHY workload whose cost scales with both workload parallelism and decoder iteration budget, making it a potential bottleneck on general-purpose central processing units (CPUs). This paper presents a reproducible, telemetry-backed microbenchmark derived from the Sionna LDPC5G baseline to characterize the compute headroom obtained through graphics processing unit (GPU) offload on compact heterogeneous edge platforms. We evaluate decoder behavior across multiple processor architectures and a wide range of batch sizes and iteration counts, with emphasis on dense operating regimes relevant to edge provisioning. Results show that GPU acceleration substantially increases LDPC throughput, reduces amortized decode service time, and shifts compute pressure away from the CPU, thereby improving the feasibi
This paper investigates the joint optimization of condition-based maintenance and spare provisioning, incorporating insights obtained from sensor data. Prognostic models estimate components' remaining lifetime distributions (RLDs), which are integrated into an optimization model to coordinate maintenance and spare provisioning. The existing literature addressing this problem assumes that prognostic models provide accurate estimates of RLDs, thereby allowing a direct adoption of Stochastic Programming or Markov Decision Process methodologies. Nevertheless, this assumption often does not hold in practice since the estimated distributions can be inaccurate due to noisy sensors or scarcity of training data. To tackle this issue, we develop a Distributionally Robust Chance Constrained (DRCC) formulation considering general discrepancy-based ambiguity sets that capture potential distribution perturbations of the estimated RLDs. The proposed formulation admits a Mixed-Integer Linear Programming (MILP) reformulation, where explicit formulas are provided to simplify the general discrepancy-based ambiguity sets. Finally, for the numerical illustration, we test a type-$\infty$ Wasserstein amb
Vision-language models (VLMs) work well in tasks ranging from image captioning to visual question answering (VQA), yet they struggle with spatial reasoning, a key skill for understanding our physical world that humans excel at. We find that spatial relations are generally rare in widely used VL datasets, with only a few being well represented, while most form a long tail of underrepresented relations. This gap leaves VLMs ill-equipped to handle diverse spatial relationships. To bridge it, we construct a synthetic VQA dataset focused on spatial reasoning generated from hyper-detailed image descriptions in Localized Narratives, DOCCI, and PixMo-Cap. Our dataset consists of 455k samples containing 3.4 million QA pairs. Trained on this dataset, our Spatial-Reasoning Enhanced (SpaRE) VLMs show strong improvements on spatial reasoning benchmarks, achieving up to a 49% performance gain on the What's Up benchmark, while maintaining strong results on general tasks. Our work narrows the gap between human and VLM spatial reasoning and makes VLMs more capable in real-world tasks such as robotics and navigation.
Restoring limb motor function in individuals with spinal cord injury (SCI), stroke, or amputation remains a critical challenge, one which affects millions worldwide. Recent studies show through surface electromyography (EMG) that spared motor neurons can still be voluntarily controlled, even without visible limb movement . These signals can be decoded and used for motor intent estimation; however, current wearable solutions lack the necessary hardware and software for intuitive interfacing of the spared degrees of freedom after neural injuries. To address these limitations, we developed a wireless, high-density EMG bracelet, coupled with a novel software framework, MyoGestic. Our system allows rapid and tailored adaptability of machine learning models to the needs of the users, facilitating real-time decoding of multiple spared distinctive degrees of freedom. In our study, we successfully decoded the motor intent from two participants with SCI, two with spinal stroke , and three amputees in real-time, achieving several controllable degrees of freedom within minutes after wearing the EMG bracelet. We provide a proof-of-concept that these decoded signals can be used to control a digi
This paper introduces a Markov chain-based approach for the analysis and optimization of spare-management policies in large-scale satellite constellations. Focusing on the direct strategy, we model spare replenishment as a periodic-review reorder-point/order-quantity policy, where spares are deployed directly to constellation planes. The stochastic behavior of satellite failures and launch vehicle lead times is captured through Markov representations of both failure and replenishment dynamics. Based on this efficient and accurate framework, we construct and solve an optimization problem aimed at minimizing operational costs. The effectiveness of the proposed method is demonstrated through a case study using a real-world mega-constellation.
While there has been extensive work on deep neural networks for images and text, deep learning for relational databases (RDBs) is still a rather unexplored field. One direction that recently gained traction is to apply Graph Neural Networks (GNNs) to RBDs. However, training GNNs on large relational databases (i.e., data stored in multiple database tables) is rather inefficient due to multiple rounds of training and potentially large and inefficient representations. Hence, in this paper we propose SPARE (Single-Pass Relational models), a new class of neural models that can be trained efficiently on RDBs while providing similar accuracies as GNNs. For enabling efficient training, different from GNNs, SPARE makes use of the fact that data in RDBs has a regular structure, which allows one to train these models in a single pass while exploiting symmetries at the same time. Our extensive empirical evaluation demonstrates that SPARE can significantly speedup both training and inference while offering competitive predictive performance over numerous baselines.
We study a repairable inventory system dedicated to a single component that is critical in operating a capital good. The system consists of a stock point containing spare components, and a dedicated repair shop responsible for repairing damaged components. Components are replaced using an age-replacement strategy, which sends components to the repair shop either preventively if it reaches the age-threshold, and correctively otherwise. Damaged components are replaced by new ones if there are spare components available, otherwise the capital good is inoperable. If there is free capacity in the repair shop, then the repair of the damaged component immediately starts, otherwise it is queued. The manager decides on the number of repairables in the system, the age-threshold, and the capacity of the repair shop. There is an inherent trade-off: A low (high) age-threshold reduces (increases) the probability of a corrective replacement but increases (decreases) the demand for repair capacity, and a high (low) number of repairables in the system leads to higher (lower) holding costs, but decreases (increases) the probability of downtime. We first show that the single capital good setting can
The recent growing trend to develop large-scale satellite constellations (i.e., mega-constellation) with low-cost small satellites has brought the need for an efficient and scalable maintenance strategy decision plan. Traditional spare strategies for satellite constellations cannot handle these mega-constellations due to their limited scalability in number of satellites and/or frequency of failures. In this paper, we propose a novel spare strategy using an inventory management approach. We consider a set of parking orbits at a lower altitude than the constellation for spare storage, and model satellite constellation spare strategy problem using a multi-echelon (s,Q)-type inventory policy, viewing Earth's ground as a supplier, parking orbits as warehouses, and in-plane spare stocks as retailers. This inventory model is unique in that the parking orbits (warehouses) drift away from the orbital planes over time due to orbital mechanics' effects, and the in-plane spare stocks (retailers) would receive the resupply from the closest (i.e., minimum waiting time) available warehouse at the time of delivery. The parking orbits (warehouses) are also resupplied from the ground (supplier) with
Existing optimization-based methods for non-rigid registration typically minimize an alignment error metric based on the point-to-point or point-to-plane distance between corresponding point pairs on the source surface and target surface. However, these metrics can result in slow convergence or a loss of detail. In this paper, we propose SPARE, a novel formulation that utilizes a symmetrized point-to-plane distance for robust non-rigid registration. The symmetrized point-to-plane distance relies on both the positions and normals of the corresponding points, resulting in a more accurate approximation of the underlying geometry and can achieve higher accuracy than existing methods. To solve this optimization problem efficiently, we introduce an as-rigid-as-possible regulation term to estimate the deformed normals and propose an alternating minimization solver using a majorization-minimization strategy. Moreover, for effective initialization of the solver, we incorporate a deformation graph-based coarse alignment that improves registration quality and efficiency. Extensive experiments show that the proposed method greatly improves the accuracy of non-rigid registration problems and ma
Non-semantic features or semantic-agnostic features, which are irrelevant to image context but sensitive to image manipulations, are recognized as evidential to Image Manipulation Localization (IML). Since manual labels are impossible, existing works rely on handcrafted methods to extract non-semantic features. Handcrafted non-semantic features jeopardize IML model's generalization ability in unseen or complex scenarios. Therefore, for IML, the elephant in the room is: How to adaptively extract non-semantic features? Non-semantic features are context-irrelevant and manipulation-sensitive. That is, within an image, they are consistent across patches unless manipulation occurs. Then, spare and discrete interactions among image patches are sufficient for extracting non-semantic features. However, image semantics vary drastically on different patches, requiring dense and continuous interactions among image patches for learning semantic representations. Hence, in this paper, we propose a Sparse Vision Transformer (SparseViT), which reformulates the dense, global self-attention in ViT into a sparse, discrete manner. Such sparse self-attention breaks image semantics and forces SparseViT t
Considering the close interaction between spare parts logistics and maintenance planning, this paper presents a model for joint optimization of multi-location spare parts supply chain and condition-based maintenance under predictive and opportunistic approaches. Simultaneous use of the imperfect maintenance actions and innovative policy on spare part ordering, which is defined based on the deterioration characteristic of the system, is a significant contribution to the research. This paper also proposes the method to determine the inspection time which not only considers restraints of the both maintenance and spare parts provision policies, but also uses an event-driven approach in order to prevent unnecessary inspections. Defined decision variables such reliability, upper limit for spare parts order quantity, preventive maintenance threshold, re-ordering level of degradation, and the maximum level of successive imperfect actions will be optimized via stochastic Monte-Carlo simulation. The optimization follows two objectives: (1) system should reach the expected availability which helps decision makers apply the opportunistic approach (2) and cost rate function as an objective func
Despite their remarkable performance, Vision Language Models (VLMs) incur substantial computational overhead due to the large number of visual tokens. While diversity maximization has become a dominant strategy for token reduction, existing methods rely on cosine-based normalized similarity that discards magnitude information, failing to faithfully approximate the original feature representation and leading to suboptimal performance, particularly on compositional multi-skill reasoning tasks. In this paper, we introduce SPARE, a subspace reconstruction method that reformulates token pruning as a column subset selection problem and explicitly minimizes reconstruction error. By iteratively selecting tokens with large projection residuals, SPARE performs reconstruction-driven pruning beyond angular diversity. Moreover, we reveal a counterintuitive anti-relevance phenomenon: tokens with lower image-text relevance score can better preserve contextual information. Based on this finding, we incorporate anti-relevance into SPARE as an additional selection criterion to promote context-aware token selection. Extensive experiments across multiple VLMs and benchmarks demonstrate that SPARE cons