arXiv subjects

Nicola Paoletti

Publications and source records attributed to Nicola Paoletti.

At least 19 recordsLinked to original sources

arXiv2026

Calibrate-Then-Delegate: Safety Monitoring with Risk and Budget Guarantees via Model Cascades

Monitoring LLM safety at scale requires balancing cost and accuracy: a cheap latent-space probe can screen every input, but hard cases should be escalated to a more expensive expert. Existing cascades delegate based on probe uncertainty, but uncertainty is a poor proxy for the utility of an expert call, as it ignores whether the expert would actually improve the prediction. To address this problem, we introduce Calibrate-Then-Delegate (CTD), a model-cascade approach that provides probabilistic guarantees on either computation cost or safety performance while enabling instance-level (streaming) decisions. CTD builds on a novel delegation value (DV) probe, a lightweight model operating on the same internal representations as the safety probe that predicts the benefit of replacing the probe score with the expert score after an expert call. CTD calibrates a threshold on the DV signal using held-out data and multiple hypothesis testing, yielding finite-sample guarantees while optimising the complementary objective. Evaluated on four safety datasets using learned-attention probes, CTD improves over uncertainty-based delegation over much of the budget range, avoids unnecessary expert calls, and extracts positive-value subsets even from groups where the expert score is worse than the probe score on average. Ablations across four probe architectures show that these findings are not specific to any particular architecture.

cs.LG↗

arXiv2026

Reliable Inference in Edge-Cloud Model Cascades via Conformal Alignment

Edge intelligence enables low-latency inference via compact on-device models, but assuring reliability remains challenging. We study edge-cloud cascades that must preserve conditional coverage: whenever the edge returns a prediction set, it should contain the true label with a user-specified probability, as if produced by the cloud model. We formalize conditional coverage with respect to the cloud predictive distribution, and introduce a conformal alignment-based (CAb) cascading mechanism that certifies this property with user control over the risk level. Our method casts escalation from edge to cloud models as a multiple-hypothesis testing (MHT) problem, tailoring conformal alignment (CA) to select which inputs can be safely handled at the edge. The proposed CAb model cascading method yields statistical guarantees on the average fraction of edge decisions that satisfy cloud-level conditional coverage. The procedure applies to arbitrary edge prediction sets, including variants of conformal prediction (CP), and exposes a tunable trade-off among coverage, deferral rate, and set size. Experiments on CIFAR-100 image classification and the TeleQnA question-answering (QA) benchmark show that the proposed CAb cascade maintains the target conditional coverage for edge predictions while substantially reducing offloading to the cloud and incurring modest increases in prediction-set size.

cs.LG↗

arXiv2026

Online Conformal Prediction Beyond Feedback

Uncertainty quantification is essential when deploying machine learning models in safety-critical applications. Online conformal prediction (OCP) provides theoretically principled uncertainty quantification for arbitrary black-box classifiers and non-i.i.d. data streams by constructing prediction sets that are guaranteed to contain the true label at a user-specified frequency. OCP usually updates prediction sets using feedback from previously deployed predictions. We instead study an OCP setting beyond feedback: on each round, the learner can either output a prediction set or query the correct label, but not both. Thus, no deployed prediction is ever evaluated directly. We reduce this problem to a partial monitoring game in which prediction actions return no observation and a separate query action reveals the label. The reward function is constructed in a way that encourages the learner to output small prediction sets while ensuring that the correct label is covered with a sufficiently high probability. To solve this game, we develop OCP with queries (OCPQ) by adapting the label efficient forecaster of Cesa-Bianchi, Lugosi, and Stoltz (2004) to our setting. For any black box classifier and any (non-i.i.d.) oblivious data stream of length $T$, OCPQ has $O(T^{2/3})$ expected regret and expected coverage at least $β-O(T^{-1/3})$ for a user-defined $β$, while querying only an expected $T^{-1/3}$ fraction of rounds. This provides coverage comparable to bandit-based OCP methods while requiring no feedback from deployed prediction sets. Experiments on real-world datasets further demonstrate the effectiveness of our approach.

cs.LG↗

arXiv2026

Robust Counterfactual Policy Optimisation via Nondeterministic Causal Models

Counterfactual inference approaches for sequential decision-making typically assume deterministic causal models, where all randomness stems from latent variables. However, Markov Decision Processes (MDPs) are inherently stochastic. We address this by formalising counterfactual policy optimisation under probabilistic nondeterministic causal models, which properly separates latent confounding from irreducible stochasticity, and here propose a first practical optimisation problem for identifying robust counterfactual policies under a sensitivity analysis framework. We validate our approach on a sepsis treatment simulator, where diabetes status acts as a hidden global confounder.

cs.LG↗

arXiv2026

On the QUEST for Uncertainty Quantification via Highest Density Regions

Uncertainty quantification (UQ) is essential for reliable decision-making in safety-critical applications in probabilistic machine learning. For regression problems, dominant scalar UQ approaches - notably, those based on proper scoring rules - measure uncertainty via pointwise predictive risk. This can lead to counterintuitive results when the target statistic is not the conditional expectation. We propose an alternative framework, in which uncertainty is characterised by the volume of the most probable subset of a distribution's support. QUEST (Quantifying Uncertainty via highest dEnSiTy regions) is a novel approach to UQ based on the concentration of Lebesgue measure at a distribution's peak(s), evaluated at one or more values of a robustness parameter $α$. We establish connections between our measures and classical statistics from information theory and economics. We show that, unlike popular alternatives based on proper scoring rules, QUEST measures of epistemic and aleatoric uncertainty satisfy a set of axioms adapted from the UQ literature, including monotonicity under distributional spread and invariance to location shifts. Selective prediction benchmarks confirm that QUEST performs favourably against standard measures such as variance and differential entropy.

cs.LG↗

arXiv2026

PreAct-Bench: Benchmarking Predictive Monitoring in LLMs

Large language models (LLMs) are increasingly deployed as autonomous agents capable of executing multi-step action trajectories toward a given objective. While existing safety research has focused on detecting unethical behavior from complete trajectories, this paradigm is fundamentally retrospective: it identifies harm only after it has already occurred. In this work, we study a critical yet overlooked safety task, which we term Predictive Monitoring: given only a partial action trajectory, can a model infer whether it will culminate in an unethical action before the overt action is executed? To support this task, we present PreActBench, a benchmark of 1,000 paired ethical and unethical action trajectories spanning five domains. We evaluate a range of LLMs, safety guardrail models, and latent probing methods across varying fractions of the action trajectory using our Prefix Foresight F1 metric. Results show that while humans achieve promising performance, predictive monitoring remains challenging even for strong models, highlighting the need for future-oriented risk reasoning in LLM safety.

cs.LG↗

arXiv2026

Robust Counterfactual Inference in Markov Decision Processes

This paper addresses a key limitation in existing counterfactual inference methods for Markov Decision Processes (MDPs). Current approaches assume a specific causal model to make counterfactuals identifiable. However, there are usually many causal models that align with the observational and interventional distributions of an MDP, each yielding different counterfactual distributions, so fixing a particular causal model limits the validity (and usefulness) of counterfactual inference. We propose a novel non-parametric approach that computes tight bounds on counterfactual transition probabilities across all compatible causal models. Unlike previous methods that require solving prohibitively large optimisation problems (with variables that grow exponentially in the size of the MDP), our approach provides closed-form expressions for these bounds, making computation highly efficient and scalable for non-trivial MDPs. Once such an interval counterfactual MDP is constructed, our method identifies robust counterfactual policies that optimise the worst-case reward w.r.t. the uncertain interval MDP probabilities. We evaluate our method on various case studies, demonstrating improved robustness over existing methods.

cs.AI↗

arXiv2026

Physics-Informed Neural Operators for Cardiac Electrophysiology

Accurately simulating systems governed by PDEs, such as voltage fields in cardiac electrophysiology (EP) modelling, remains a significant modelling challenge. Traditional numerical solvers are computationally expensive and sensitive to discretisation, while canonical deep learning methods are data-hungry and struggle with chaotic dynamics and long-term predictions. Physics-Informed Neural Networks (PINNs) mitigate some of these issues by incorporating physical constraints in the learning process, yet they remain limited by mesh resolution and long-term predictive stability. In this work, we propose a Physics-Informed Neural Operator (PINO) approach to solve PDE problems in cardiac EP. Unlike PINNs, PINO models learn mappings between function spaces, allowing them to generalise to multiple mesh resolutions and initial conditions. Our results show that PINO models can accurately reproduce cardiac EP dynamics over extended time horizons and across multiple propagation scenarios, including zero-shot evaluations on scenarios unseen during training. Additionally, our PINO models maintain high predictive quality in long roll-outs (where predictions are recursively fed back as inputs), and can scale their predictive resolution by up to 10x the training resolution. These advantages come with a significant reduction in simulation time compared to numerical PDE solvers, highlighting the potential of PINO-based approaches for efficient and scalable cardiac EP simulations.

cs.LG↗

arXiv2025

An STREL-based Formulation of Spatial Resilience in Cyber-Physical Systems

Resiliency is the ability of a system to quickly recover from a violation (recoverability) and avoid future violations for as long as possible (durability). In the spatial setting, recoverability and durability (now known as persistency) are measured in units of distance. Like its temporal counterpart, spatial resiliency is of fundamental importance for Cyber-Physical Systems (CPS) and yet, to date, there is no widely agreed-upon formal treatment of spatial resiliency. We present a formal framework for reasoning about spatial resiliency in CPS. Our framework is based on the spatial fragment of STREL, which we refer to as SREL. In this framework, spatial resiliency is given a syntactic characterization in the form of a Spatial Resiliency Specification (SpaRS). An atomic predicate of SpaRS is called an S-atom. Given an arbitrary SREL formula $φ$, distance bounds $d_1, d_2$, the S-atom of $φ$, $S_{d_1, d_2} (φ)$, is the SREL formula $\negφR_{[0,d_1]} (φR_{[d_2, +\infty)}φ)$, specifying that recovery from a violation of $φ$ occurs within distance $d_1$ (recoverability), and subsequently that $φ$ be maintained along a route for a distance greater than $d_2$ (persistency). S-atoms can be combined using spatial STREL operators, allowing one to express composite resiliency specifications. We define a quantitative semantics for SpaRS in the form of a Spatial Resilience Value (SpaRV) function $σ$ and prove its soundness and completeness w.r.t. SREL's Boolean semantics. The $σ$-value for $S_{d_1,d_2}(φ)$ is a set of non-dominated (rec, per) pairs, quantifying recoverability and persistency, given that some routes may offer better recoverability while others better persistency. In addition, we design algorithms to evaluate SpaRV for SpaRS formulas. Finally, two case studies demonstrate the practical utility of our approach.

cs.LO↗

arXiv2025

DRMD: Deep Reinforcement Learning for Malware Detection under Concept Drift

Malware detection in real-world settings must deal with evolving threats, limited labeling budgets, and uncertain predictions. Traditional classifiers, without additional mechanisms, struggle to maintain performance under concept drift in malware domains, as their supervised learning formulation cannot optimize when to defer decisions to manual labeling and adaptation. Modern malware detection pipelines combine classifiers with monthly active learning (AL) and rejection mechanisms to mitigate the impact of concept drift. In this work, we develop a novel formulation of malware detection as a one-step Markov Decision Process and train a deep reinforcement learning (DRL) agent, simultaneously optimizing sample classification performance and rejecting high-risk samples for manual labeling. We evaluated the joint detection and drift mitigation policy learned by the DRL-based Malware Detection (DRMD) agent through time-aware evaluations on Android malware datasets subject to realistic drift requiring multi-year performance stability. The policies learned under these conditions achieve a higher Area Under Time (AUT) performance compared to standard classification approaches used in the domain, showing improved resilience to concept drift. Specifically, the DRMD agent achieved an average AUT improvement of 8.66 and 10.90 for the classification-only and classification-rejection policies, respectively. Our results demonstrate for the first time that DRL can facilitate effective malware detection and improved resiliency to concept drift in the dynamic setting of Android malware detection.

cs.LG↗

arXiv2025

CONFEX: Uncertainty-Aware Counterfactual Explanations with Conformal Guarantees

Counterfactual explanations (CFXs) provide human-understandable justifications for model predictions, enabling actionable recourse and enhancing interpretability. To be reliable, CFXs must avoid regions of high predictive uncertainty, where explanations may be misleading or inapplicable. However, existing methods often neglect uncertainty or lack principled mechanisms for incorporating it with formal guarantees. We propose CONFEX, a novel method for generating uncertainty-aware counterfactual explanations using Conformal Prediction (CP) and Mixed-Integer Linear Programming (MILP). CONFEX explanations are designed to provide local coverage guarantees, addressing the issue that CFX generation violates exchangeability. To do so, we develop a novel localised CP procedure that enjoys an efficient MILP encoding by leveraging an offline tree-based partitioning of the input space. This way, CONFEX generates CFXs with rigorous guarantees on both predictive uncertainty and optimality. We evaluate CONFEX against state-of-the-art methods across diverse benchmarks and metrics, demonstrating that our uncertainty-aware approach yields robust and plausible explanations.

cs.LG↗

arXiv2025

Distilling Calibration via Conformalized Credal Inference

Deploying artificial intelligence (AI) models on edge devices involves a delicate balance between meeting stringent complexity constraints, such as limited memory and energy resources, and ensuring reliable performance in sensitive decision-making tasks. One way to enhance reliability is through uncertainty quantification via Bayesian inference. This approach, however, typically necessitates maintaining and running multiple models in an ensemble, which may exceed the computational limits of edge devices. This paper introduces a low-complexity methodology to address this challenge by distilling calibration information from a more complex model. In an offline phase, predictive probabilities generated by a high-complexity cloud-based model are leveraged to determine a threshold based on the typical divergence between the cloud and edge models. At run time, this threshold is used to construct credal sets -- ranges of predictive probabilities that are guaranteed, with a user-selected confidence level, to include the predictions of the cloud model. The credal sets are obtained through thresholding of a divergence measure in the simplex of predictive probabilities. Experiments on visual and language tasks demonstrate that the proposed approach, termed Conformalized Distillation for Credal Inference (CD-CI), significantly improves calibration performance compared to low-complexity Bayesian methods, such as Laplace approximation, making it a practical and efficient solution for edge AI deployments.

cs.LG↗

arXiv2025

Conformal Predictive Monitoring for Multi-Modal Scenarios

We consider the problem of quantitative predictive monitoring (QPM) of stochastic systems, i.e., predicting at runtime the degree of satisfaction of a desired temporal logic property from the current state of the system. Since computational efficiency is key to enable timely intervention against predicted violations, several state-of-the-art QPM approaches rely on fast machine-learning surrogates to provide prediction intervals for the satisfaction values, using conformal inference to offer statistical guarantees. However, these QPM methods suffer when the monitored agent exhibits multi-modal dynamics, whereby certain modes may yield high satisfaction values while others critically violate the property. Existing QPM methods are mode-agnostic and so would yield overly conservative and uninformative intervals that lack meaningful mode-specific satisfaction information. To address this problem, we present GenQPM, a method that leverages deep generative models, specifically score-based diffusion models, to reliably approximate the probabilistic and multi-modal system dynamics without requiring explicit model access. GenQPM employs a mode classifier to partition the predicted trajectories by dynamical mode. For each mode, we then apply conformal inference to produce statistically valid, mode-specific prediction intervals. We demonstrate the effectiveness of GenQPM on a benchmark of agent navigation and autonomous driving tasks, resulting in prediction intervals that are significantly more informative (less conservative) than mode-agnostic baselines.

cs.AI↗

arXiv2025

Abstract Counterfactuals for Language Model Agents

Counterfactual inference is a powerful tool for analysing and evaluating autonomous agents, but its application to language model (LM) agents remains challenging. Existing work on counterfactuals in LMs has primarily focused on token-level counterfactuals, which are often inadequate for LM agents due to their open-ended action spaces. Unlike traditional agents with fixed, clearly defined action spaces, the actions of LM agents are often implicit in the strings they output, making their action spaces difficult to define and interpret. Furthermore, the meanings of individual tokens can shift depending on the context, adding complexity to token-level reasoning and sometimes leading to biased or meaningless counterfactuals. We introduce \emph{Abstract Counterfactuals}, a framework that emphasises high-level characteristics of actions and interactions within an environment, enabling counterfactual reasoning tailored to user-relevant features. Our experiments demonstrate that the approach produces consistent and meaningful counterfactuals while minimising the undesired side effects of token-level methods. We conduct experiments on text-based games and counterfactual text generation, while considering both token-level and latent-space interventions.

cs.LG↗

arXiv2025

SimICD: A Closed-Loop Simulation Framework For ICD Therapy

Virtual studies of ICD behaviour are crucial for testing device functionality in a controlled environment prior to clinical application. Although previous works have shown the viability of using in silico testing for diagnosis, there is a notable gap in available models that can simulate therapy progression decisions during arrhythmic episodes. This work introduces SimICD, a simulation tool which combines virtual ICD logic algorithms with cardiac electrophysiology simulations in a feedback loop, allowing the progression of ICD therapy protocols to be simulated for a range of tachy-arrhythmia episodes. Using a cohort of virtual patients, we demonstrate the ability of SimICD to simulate realistic cardiac signals and ICD responses that align with the logic of real-world devices, facilitating the reprogramming of ICD parameters to adapt to specific episodes.

cs.CE↗

arXiv2025

Counterfactual Influence in Markov Decision Processes

Our work addresses a fundamental problem in the context of counterfactual inference for Markov Decision Processes (MDPs). Given an MDP path $τ$, this kind of inference allows us to derive counterfactual paths $τ'$ describing what-if versions of $τ$ obtained under different action sequences than those observed in $τ$. However, as the counterfactual states and actions deviate from the observed ones over time, the observation $τ$ may no longer influence the counterfactual world, meaning that the analysis is no longer tailored to the individual observation, resulting in interventional outcomes rather than counterfactual ones. Even though this issue specifically affects the popular Gumbel-max structural causal model used for MDP counterfactuals, it has remained overlooked until now. In this work, we introduce a formal characterisation of influence based on comparing counterfactual and interventional distributions. We devise an algorithm to construct counterfactual models that automatically satisfy influence constraints. Leveraging such models, we derive counterfactual policies that are not just optimal for a given reward structure but also remain tailored to the observed path. Even though there is an unavoidable trade-off between policy optimality and strength of influence constraints, our experiments demonstrate that it is possible to derive (near-)optimal policies while remaining under the influence of the observation.

cs.AI↗

arXiv2025

LTL Verification of Memoryful Neural Agents

We present a framework for verifying Memoryful Neural Multi-Agent Systems (MN-MAS) against full Linear Temporal Logic (LTL) specifications. In MN-MAS, agents interact with a non-deterministic, partially observable environment. Examples of MN-MAS include multi-agent systems based on feed-forward and recurrent neural networks or state-space models. Different from previous approaches, we support the verification of both bounded and unbounded LTL specifications. We leverage well-established bounded model checking techniques, including lasso search and invariant synthesis, to reduce the verification problem to that of constraint solving. To solve these constraints, we develop efficient methods based on bound propagation, mixed-integer linear programming, and adaptive splitting. We evaluate the effectiveness of our algorithms in single and multi-agent environments from the Gymnasium and PettingZoo libraries, verifying unbounded specifications for the first time and improving the verification time for bounded specifications by an order of magnitude compared to the SoA.

cs.LO↗

arXiv2025

Certified Guidance for Planning with Deep Generative Models

Deep generative models, such as generative adversarial networks and diffusion models, have recently emerged as powerful tools for planning tasks and behavior synthesis in autonomous systems. Various guidance strategies have been introduced to steer the generative process toward outputs that are more likely to satisfy the planning objectives. These strategies avoid the need for model retraining but do not provide any guarantee that the generated outputs will satisfy the desired planning objectives. To address this limitation, we introduce certified guidance, an approach that modifies a generative model, without retraining it, into a new model guaranteed to satisfy a given specification with probability one. We focus on Signal Temporal Logic specifications, which are rich enough to describe nontrivial planning tasks. Our approach leverages neural network verification techniques to systematically explore the latent spaces of the generative models, identifying latent regions that are certifiably correct with respect to the STL property of interest. We evaluate the effectiveness of our method on four planning benchmarks using GANs and diffusion models. Our results confirm that certified guidance produces generative models that are always correct, unlike existing guidance methods that are not certified.

cs.LG↗