Author name cluster

Luca Oneto

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

13 papers

2 author rows

AAAI Conference 2026 Conference Paper

SOM Directions Are Better than One: Multi-Directional Refusal Suppression in Language Models

Giorgio Piras
Raffaele Mura
Fabio Brau
Luca Oneto
Fabio Roli
Battista Biggio

Refusal refers to the functional behavior enabling safety-aligned language models to reject harmful or unethical prompts. Following the growing scientific interest in mechanistic interpretability, recent work encoded refusal behavior as a single direction in the model’s latent space; e.g., computed as the difference between the centroids of harmful and harmless prompt representations. However, emerging evidence suggests that concepts in LLMs often appear to be encoded as a low-dimensional manifold embedded in the high-dimensional latent space. Motivated by these findings, we propose a novel method leveraging Self-Organizing Maps (SOMs) to extract multiple refusal directions. To this end, we first prove that SOMs generalize the prior work's difference-in-means technique. We then train SOMs on harmful prompt representations to identify multiple neurons. By subtracting the centroid of harmless representations from each neuron, we derive a set of multiple directions expressing the refusal concept. We validate our method on an extensive experimental setup, demonstrating that ablating multiple directions from models' internals outperforms not only the single-direction baseline but also specialized jailbreak algorithms, leading to an effective suppression of refusal. Finally, we conclude by analyzing the mechanistic implications of our approach.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

Training-Free Constrained Generation With Stable Diffusion Models

Stefano Zampini
Jacob K Christopher
Luca Oneto
Davide Anguita
Ferdinando Fioretto

Stable diffusion models represent the state-of-the-art in data synthesis across diverse domains and hold transformative potential for applications in science and engineering, e. g. , by facilitating the discovery of novel solutions and simulating systems that are computationally intractable to model explicitly. While there is increasing effort to incorporate physics-based constraints into generative models, existing techniques are either limited in their applicability to latent diffusion frameworks or lack the capability to strictly enforce domain-specific constraints. To address this limitation this paper proposes a novel integration of stable diffusion models with constrained optimization frameworks, enabling the generation of outputs satisfying stringent physical and functional requirements. The effectiveness of this approach is demonstrated through material design experiments requiring adherence to precise morphometric properties, challenging inverse design tasks involving the generation of materials inducing specific stress-strain responses, and copyright-constrained content generation tasks. All code has been released at https: //github. com/RAISELab-atUVA/Constrained-Stable-Diffusion.

PDF Details

NeurIPS Conference 2025 Conference Paper

TransferBench: Benchmarking Ensemble-based Black-box Transfer Attacks

Fabio Brau
Maura Pintor
Antonio Cinà
Raffaele Mura
Luca Scionis
Luca Oneto
Fabio Roli
Battista Biggio

Ensemble-based black-box transfer attacks optimize adversarial examples on a set of surrogate models, claiming to reach high success rates by querying the (unknown) target model only a few times. In this work, we show that prior evaluations are systematically biased, as such methods are tested only under overly optimistic scenarios, without considering (i) how the choice of surrogate models influences transferability, (ii) how they perform against robust target models, and (iii) whether querying the target to refine the attack is really required. To address these gaps, we introduce TransferBench, a framework for evaluating ensemble-based black-box transfer attacks under more realistic and challenging scenarios than prior work. Our framework considers 17 distinct settings on CIFAR-10 and ImageNet, including diverse surrogate-target combinations, robust targets, and comparisons to baseline methods that do not use any query-based refinement mechanism. Our findings reveal that existing methods fail to generalize to more challenging scenarios, and that query-based refinement offers little to no benefit, contradicting prior claims. These results highlight that building reliable and query-efficient black-box transfer attacks remains an open challenge. We release our benchmark and evaluation code at: https: //github. com/pralab/transfer-bench.

PDF Details

EAAI Journal 2023 Journal Article

Physically plausible propeller noise prediction via recursive corrections leveraging prior knowledge and experimental data

Miltiadis Kalikatzarakis
Andrea Coraddu
Mehmet Atlar
Stefano Gaggero
Giorgio Tani
Luca Oneto

For propeller-driven vessels, cavitation is the most dominant noise source producing both structure-borne and radiated noise impacting wildlife, passenger comfort, and underwater warfare. Physically plausible and accurate predictions of the underwater radiated noise at design stage, i. e. , for previously untested geometries and operating conditions, are fundamental for designing silent and efficient propellers. State-of-the-art predictive models are based on physical, data-driven, and hybrid approaches. Physical models (PMs) meet the need for physically plausible predictions but are either too computationally demanding or not accurate enough at design stage. Data-driven models (DDMs) are computationally inexpensive ad accurate on average but sometimes produce physically implausible results. Hybrid models (HMs) combine PMs and DDMs trying to take advantage of their strengths while limiting their weaknesses but state-of-the-art hybridisation strategies do not actually blend them, failing to achieve the HMs full potential. In this work, for the first time, we propose a novel HM that recursively correct a state-of-the-art PM by means of a DDM which simultaneously exploits the prior physical knowledge in the definition of its feature set and the data coming from a vast experimental campaign at the Emerson Cavitation Tunnel on the Meridian standard propeller series behind different severities of the axial wake. Results in different extrapolating conditions, i. e. , extrapolation with respect to propeller rotational speed, wakefield, and geometry, will support our proposal both in terms of accuracy and physical plausibility.

Details DOI

SoCS Conference 2021 Conference Paper

A Planning-based Approach for In-Station Train Dispatching

Matteo Cardellini
Marco Maratea
Mauro Vallati
Gianluca Boleto
Luca Oneto

In-station train dispatching is the problem of optimising the effective utilisation of available railway infrastructures for mitigating incidents and delays. In this paper, we describe an approach for dealing with the in-station dispatching problem by means of automated planning techniques.

Details

ICAPS Conference 2021 Conference Paper

In-Station Train Dispatching: A PDDL+ Planning Approach

Matteo Cardellini
Marco Maratea
Mauro Vallati
Gianluca Boleto
Luca Oneto

In railway networks, stations are probably the most critical points for interconnecting trains' routes: in a restricted geographical area, a potentially large number of trains have to stop according to an official timetable, with the concrete risk of accumulating delays that can then have a knockout effect on the rest of the network. In this context, in-station train dispatching plays a central role in maximising the effective utilisation of available railway infrastructures and in mitigating the impact of incidents and delays. Unfortunately, in-station train dispatching is still largely handled manually by human operators in charge of a group of stations. In this paper we make a step towards supporting the operator with some automatic tool, by describing an approach for performing in-station dispatching by means of automated planning techniques. Given the mixed discrete-continuous nature of the problem, we employ PDDL+ for the specification of the problem, and the ENHSP planning engine enhanced by domain-specific solving techniques. Results on a range of scenarios, using real-data of a station of the North West of Italy, show the potential of our approach.

Details

EAAI Journal 2021 Journal Article

Marine dual fuel engines monitoring in the wild through weakly supervised data analytics

Andrea Coraddu
Luca Oneto
Davide Ilardi
Sokratis Stoumpos
Gerasimos Theotokatos

Background: Maritime transportation accounts for around 80% of the world freight movements, remarkably contributing to the global environmental footprint. Dual fuel engines, running on both gaseous and liquid fuels, represent a viable way toward the reduction of emissions at the cost of additional complexity in monitoring activities. Motivation: Data-driven methods represent the frontier in research and in maritime industrial applications, and they usually require a large amount of labelled data, i. e. , sensor measurements plus the associated engine status usually annotated by human operators, which are costly and seldomly available in the wild. Unlabelled samples, instead, are commonly, cheaply, and readily available. Hypothesis: The enabling technology for data-driven methods is the availability of a network of sensors and an automation system able to capture and store the associated stream of data. Methods: In this paper, we design and propose multiple alternatives toward the weakly supervised marine dual fuel engines data-driven monitoring. To this aim, we will rely on a Digital Twin of the dual fuel engine or on novelty detection algorithms and we will compare them against state-of-the-art fully supervised approaches. Results: Results on data generated from a real-data validated simulator of a marine dual fuel engine demonstrate that the proposed weakly supervised monitoring approaches lead to a negligible loss in accuracy compared to costly and often unfeasible fully supervised ones supporting the validity of the proposal for its application in the wild. Conclusion: The main outcome is a guideline for selecting the best data-driven dual fuel engine monitoring method according to the available data.

Details DOI

NeurIPS Conference 2020 Conference Paper

Exploiting MMD and Sinkhorn Divergences for Fair and Transferable Representation Learning

Luca Oneto
Michele Donini
Giulia Luise
Carlo Ciliberto
Andreas Maurer
Massimiliano Pontil

Developing learning methods which do not discriminate subgroups in the population is a central goal of algorithmic fairness. One way to reach this goal is by modifying the data representation in order to meet certain fairness constraints. In this work we measure fairness according to demographic parity. This requires the probability of the possible model decisions to be independent of the sensitive information. We argue that the goal of imposing demographic parity can be substantially facilitated within a multitask learning setting. We present a method for learning a shared fair representation across multiple tasks, by means of different new constraints based on MMD and Sinkhorn Divergences. We derive learning bounds establishing that the learned representation transfers well to novel tasks. We present experiments on three real world datasets, showing that the proposed method outperforms state-of-the-art approaches by a significant margin.

PDF Details

NeurIPS Conference 2020 Conference Paper

Fair regression via plug-in estimator and recalibration with statistical guarantees

Evgenii Chzhen
Christophe Denis
Mohamed Hebiri
Luca Oneto
Massimiliano Pontil

We study the problem of learning an optimal regression function subject to a fairness constraint. It requires that, conditionally on the sensitive feature, the distribution of the function output remains the same. This constraint naturally extends the notion of demographic parity, often used in classification, to the regression setting. We tackle this problem by leveraging on a proxy-discretized version, for which we derive an explicit expression of the optimal fair predictor. This result naturally suggests a two stage approach, in which we first estimate the (unconstrained) regression function from a set of labeled data and then we recalibrate it with another set of unlabeled data. The recalibration step can be efficiently performed via a smooth optimization. We derive rates of convergence of the proposed estimator to the optimal fair predictor both in terms of the risk and fairness constraint. Finally, we present numerical experiments illustrating that the proposed method is often superior or competitive with state-of-the-art methods.

PDF Details

NeurIPS Conference 2020 Conference Paper

Fair regression with Wasserstein barycenters

Evgenii Chzhen
Christophe Denis
Mohamed Hebiri
Luca Oneto
Massimiliano Pontil

We study the problem of learning a real-valued function that satisfies the Demographic Parity constraint. It demands the distribution of the predicted output to be independent of the sensitive attribute. We consider the case that the sensitive attribute is available for prediction. We establish a connection between fair regression and optimal transport theory, based on which we derive a close form expression for the optimal fair predictor. Specifically, we show that the distribution of this optimum is the Wasserstein barycenter of the distributions induced by the standard regression function on the sensitive groups. This result offers an intuitive interpretation of the optimal fair prediction and suggests a simple post-processing algorithm to achieve fairness. We establish risk and distribution-free fairness guarantees for this procedure. Numerical experiments indicate that our method is very effective in learning fair models, with a relative increase in error rate that is inferior to the relative gain in fairness.

PDF Details

NeurIPS Conference 2019 Conference Paper

Leveraging Labeled and Unlabeled Data for Consistent Fair Binary Classification

Evgenii Chzhen
Christophe Denis
Mohamed Hebiri
Luca Oneto
Massimiliano Pontil

We study the problem of fair binary classification using the notion of Equal Opportunity. It requires the true positive rate to distribute equally across the sensitive groups. Within this setting we show that the fair optimal classifier is obtained by recalibrating the Bayes classifier by a group-dependent threshold. We provide a constructive expression for the threshold. This result motivates us to devise a plug-in classification procedure based on both unlabeled and labeled datasets. While the latter is used to learn the output conditional probability, the former is used for calibration. The overall procedure can be computed in polynomial time and it is shown to be statistically consistent both in terms of the classification error and fairness measure. Finally, we present numerical experiments which indicate that our method is often superior or competitive with the state-of-the-art methods on benchmark datasets.

PDF Details

NeurIPS Conference 2018 Conference Paper

Empirical Risk Minimization Under Fairness Constraints

Michele Donini
Luca Oneto
Shai Ben-David
John Shawe-Taylor
Massimiliano Pontil

We address the problem of algorithmic fairness: ensuring that sensitive information does not unfairly influence the outcome of a classifier. We present an approach based on empirical risk minimization, which incorporates a fairness constraint into the learning problem. It encourages the conditional risk of the learned classifier to be approximately constant with respect to the sensitive variable. We derive both risk and fairness bounds that support the statistical consistency of our methodology. We specify our approach to kernel methods and observe that the fairness requirement implies an orthogonality constraint which can be easily added to these methods. We further observe that for linear models the constraint translates into a simple data preprocessing step. Experiments indicate that the method is empirically effective and performs favorably against state-of-the-art approaches.

PDF Details

NeurIPS Conference 2011 Conference Paper

The Impact of Unlabeled Patterns in Rademacher Complexity Theory for Kernel Classifiers

Luca Oneto
Davide Anguita
Alessandro Ghio
Sandro Ridella

We derive here new generalization bounds, based on Rademacher Complexity theory, for model selection and error estimation of linear (kernel) classifiers, which exploit the availability of unlabeled samples. In particular, two results are obtained: the first one shows that, using the unlabeled samples, the confidence term of the conventional bound can be reduced by a factor of three; the second one shows that the unlabeled samples can be used to obtain much tighter bounds, by building localized versions of the hypothesis class containing the optimal classifier.

PDF Details