On-Device Fine-Tuning via Backprop-Free Zeroth-Order Optimization

On-device fine-tuning is a critical capability for edge AI systems, which must support adaptation to different agentic tasks under stringent memory constraints. Conventional backpropagation (BP)-based training requires storing layer activations and optimizer states, a demand that can be only partially alleviated through checkpointing. In edge deployments in which the model weights must reside entirely in device memory, this overhead severely limits the maximum model size that can be deployed. Memory-efficient zeroth-order optimization (MeZO) alleviates this bottleneck by estimating gradients using forward evaluations alone, eliminating the need for storing intermediate activations or optimizer states. This enables significantly larger models to fit within on-chip memory, albeit at the cost of potentially longer fine-tuning wall-clock time. This paper first provides a theoretical estimate of the relative model sizes that can be accommodated under BP and MeZO training. We then numerically validate the analysis, demonstrating that MeZO exhibits accuracy advantages under on-device memory constraints, provided sufficient wall-clock time is available for fine-tuning.

Online Learning of Modular Bayesian Deep Receivers: Single-Step Adaptation with Streaming Data

Deep neural network (DNN)-based receivers offer a powerful alternative to classical model-based designs for wireless communication, especially in complex and nonlinear propagation environments. However, their adoption is challenged by the rapid variability of wireless channels, which makes pre-trained static DNN-based receivers ineffective, and by the latency and computational burden of online stochastic gradient descent (SGD)-based learning. In this work, we propose an online learning framework that enables rapid low-complexity adaptation of DNN-based receivers. Our approach is based on two main tenets. First, we cast online learning as Bayesian tracking in parameter space, enabling a single-step adaptation, which deviates from multi-epoch SGD . Second, we focus on modular DNN architectures that enable parallel, online, and localized variational Bayesian updates. Simulations with practical communication channels demonstrate that our proposed online learning framework can maintain a low error rate with markedly reduced update latency and increased robustness to channel dynamics as compared to traditional gradient descent based method.

Distributed Associative Memory via Online Convex Optimization

An associative memory (AM) enables cue-response recall, and associative memorization has recently been noted to underlie the operation of modern neural architectures such as Transformers. This work addresses a distributed setting where agents maintain a local AM to recall their own associations as well as selective information from others. Specifically, we introduce a distributed online gradient descent method that optimizes local AMs at different agents through communication over routing trees. Our theoretical analysis establishes sublinear regret guarantees, and experiments demonstrate that the proposed protocol consistently outperforms existing online optimization baselines.

Multi-Fidelity Hybrid Reinforcement Learning via Information Gain Maximization

Optimizing a reinforcement learning (RL) policy typically requires extensive interactions with a high-fidelity simulator of the environment, which are often costly or impractical. Offline RL addresses this problem by allowing training from pre-collected data, but its effectiveness is strongly constrained by the size and quality of the dataset. Hybrid offline-online RL leverages both offline data and interactions with a single simulator of the environment. In many real-world scenarios, however, multiple simulators with varying levels of fidelity and computational cost are available. In this work, we study multi-fidelity hybrid RL for policy optimization under a fixed cost budget. We introduce multi-fidelity hybrid RL via information gain maximization (MF-HRL-IGM), a hybrid offline-online RL algorithm that implements fidelity selection based on information gain maximization through a bootstrapping approach. Theoretical analysis establishes the no-regret property of MF-HRL-IGM, while empirical evaluations demonstrate its superior performance compared to existing benchmarks.

Synthetic Counterfactual Labels for Efficient Conformal Counterfactual Inference

This work addresses the problem of constructing reliable prediction intervals for individual counterfactual outcomes. Existing conformal counterfactual inference (CCI) methods provide marginal coverage guarantees but often produce overly conservative intervals, particularly under treatment imbalance when counterfactual samples are scarce. We introduce synthetic data-powered CCI (SP-CCI), a new framework that augments the calibration set with synthetic counterfactual labels generated by a pre-trained counterfactual model. To ensure validity, SP-CCI incorporates synthetic samples into a conformal calibration procedure based on risk-controlling prediction sets (RCPS) with a debiasing step informed by prediction-powered inference (PPI). We prove that SP-CCI achieves tighter prediction intervals while preserving marginal coverage, with theoretical guarantees under both exact and approximate importance weighting. Empirical results on different datasets confirm that SP-CCI consistently reduces interval width compared to standard CCI across all settings.

Data-Efficient Prediction-Powered Calibration via Cross-Validation

Calibration data are necessary to formally quantify the uncertainty of the decisions produced by an existing artificial intelligence (AI) model. To overcome the common issue of scarce calibration data, a promising approach is to employ synthetic labels produced by a (generally different) predictive model. However, fine-tuning the label-generating predictor on the inference task of interest, as well as estimating the residual bias of the synthetic labels, demand additional data, potentially exacerbating the calibration data scarcity problem. This paper introduces a novel approach that efficiently utilizes limited calibration data to simultaneously fine-tune a predictor and estimate the bias of the synthetic labels. The proposed method yields prediction sets with rigorous coverage guarantees for AI-generated decisions. Experimental results on an indoor localization problem validate the effectiveness and performance gains of our solution.

Adaptive Bayesian Single-Shot Quantum Sensing

Quantum sensing harnesses the unique properties of quantum systems to enable precision measurements of physical quantities such as time, magnetic and electric fields, acceleration, and gravitational gradients well beyond the limits of classical sensors. However, identifying suitable sensing probes and measurement schemes can be a classically intractable task, as it requires optimizing over Hilbert spaces of high dimension. In variational quantum sensing, a probe quantum system is generated via a parameterized quantum circuit (PQC), exposed to an unknown physical parameter through a quantum channel, and measured to collect classical data. PQCs and measurements are typically optimized using offline strategies based on frequentist learning criteria. This paper introduces an adaptive protocol that uses Bayesian inference to optimize the sensing policy via the maximization of the active information gain. The proposed variational methodology is tailored for non-asymptotic regimes where a single probe can be deployed in each time step, and is extended to support the fusion of estimates from multiple quantum sensing agents.

How to Bridge the Sim-to-Real Gap in Digital Twin-Aided Telecommunication Networks

Training effective artificial intelligence models for telecommunications is challenging due to the scarcity of deployment-specific data. Real data collection is expensive, and available datasets often fail to capture the unique operational conditions and contextual variability of the network environment. Digital twinning provides a potential solution to this problem, as simulators tailored to the current network deployment can generate site-specific data to augment the available training datasets. However, there is a need to develop solutions to bridge the inherent simulation-to-reality (sim-to-real) gap between synthetic and real-world data. This paper reviews recent advances on two complementary strategies: 1) the calibration of digital twins (DTs) through real-world measurements, and 2) the use of sim-to-real gap-aware training strategies to robustly handle residual discrepancies between digital twin-generated and real data. For the latter, we evaluate two conceptually distinct methods that model the sim-to-real gap either at the level of the environment via Bayesian learning or at the level of the training loss via prediction-powered inference.

Stochastic Quantum Spiking Neural Networks with Quantum Memory and Local Learning

Neuromorphic and quantum computing have recently emerged as promising paradigms for advancing artificial intelligence, each offering complementary strengths. Neuromorphic systems built on spiking neurons excel at processing time-series data efficiently through sparse, event-driven computation, consuming energy only upon input events. Quantum computing, on the other hand, leverages superposition and entanglement to explore feature spaces that are exponentially large in the number of qubits. Hybrid approaches combining these paradigms have begun to show potential, but existing quantum spiking models have important limitations. Notably, prior quantum spiking neuron implementations rely on classical memory mechanisms on single qubits, requiring repeated measurements to estimate firing probabilities, and they use conventional backpropagation on classical simulators for training. Here we propose a stochastic quantum spiking (SQS) neuron model that addresses these challenges. The SQS neuron uses multi-qubit quantum circuits to realize a spiking unit with internal quantum memory, enabling event-driven probabilistic spike generation in a single shot. Furthermore, we outline how networks of SQS neurons -- dubbed SQS neural networks (SQSNNs) -- can be trained via a hardware-friendly local learning rule, eliminating the need for global classical backpropagation. The proposed SQSNN model fuses the time-series efficiency of neuromorphic computing with the exponentially large inner state space of quantum computing, paving the way for quantum spiking neural networks that are modular, scalable, and trainable on quantum hardware.

Conservative quantum offline model-based optimization

Offline model-based optimization (MBO) refers to the task of optimizing a black-box objective function using only a fixed set of prior input-output data, without any active experimentation. Recent work has introduced quantum extremal learning (QEL), which leverages the expressive power of variational quantum circuits to learn accurate surrogate functions by training on a few data points. However, as widely studied in the classical machine learning literature, predictive models may incorrectly extrapolate objective values in unexplored regions, leading to the selection of overly optimistic solutions. In this paper, we propose integrating QEL with conservative objective models (COM) - a regularization technique aimed at ensuring cautious predictions on out-of-distribution inputs. The resulting hybrid algorithm, COM-QEL, builds on the expressive power of quantum neural networks while safeguarding generalization via conservative modeling. Empirical results on benchmark optimization tasks demonstrate that COM-QEL reliably finds solutions with higher true objective values compared to the original QEL, validating its superiority for offline design problems.