Stephen Hailes
YOU?
Author Swipe
View article: Opponent Shaping in LLM Agents
Opponent Shaping in LLM Agents Open
Large Language Models (LLMs) are increasingly being deployed as autonomous agents in real-world environments. As these deployments scale, multi-agent interactions become inevitable, making it essential to understand strategic behavior in s…
View article: A Generalized Information Bottleneck Theory of Deep Learning
A Generalized Information Bottleneck Theory of Deep Learning Open
The Information Bottleneck (IB) principle offers a compelling theoretical framework to understand how neural networks (NNs) learn. However, its practical utility has been constrained by unresolved theoretical ambiguities and significant ch…
View article: 20 Years in Life of a Smart Building: A retrospective
20 Years in Life of a Smart Building: A retrospective Open
Operating an intelligent smart building automation system in 2025 is met with many challenges: hardware failures, vendor obsolescence, evolving security threats and more. None of these have been comprehensibly addressed by the industrial b…
View article: Feature Selection for Network Intrusion Detection
Feature Selection for Network Intrusion Detection Open
Network Intrusion Detection (NID) remains a key area of research within the information security community, while also being relevant to Machine Learning (ML) practitioners. The latter generally aim to detect attacks using network features…
View article: Tree search in DAG space with model-based reinforcement learning for causal discovery
Tree search in DAG space with model-based reinforcement learning for causal discovery Open
Identifying causal structure is central to many fields ranging from strategic decision making to biology and economics. In this work, we propose Causal Discovery Upper Confidence Bound for Trees (CD-UCT), a model-based reinforcement learni…
View article: Socio‐Technical Security Modelling and Simulations in Cyber‐Physical Systems: Outlook on Knowledge, Perceptions, Practices, Enablers, and Barriers
Socio‐Technical Security Modelling and Simulations in Cyber‐Physical Systems: Outlook on Knowledge, Perceptions, Practices, Enablers, and Barriers Open
Socio‐Technical Security Modelling and Simulation (STSec‐M&S) is a technique used for reasoning and representing security viewpoints that include both the social and technical aspects of a system. It has shown great potential for improving…
View article: Feature Selection for Network Intrusion Detection
Feature Selection for Network Intrusion Detection Open
Network Intrusion Detection (NID) remains a key area of research within the information security community, while also being relevant to Machine Learning (ML) practitioners. The latter generally aim to detect attacks using network features…
View article: Mutual Information Preserving Neural Network Pruning
Mutual Information Preserving Neural Network Pruning Open
Pruning has emerged as the primary approach used to limit the resource requirements of large neural networks (NNs). Since the proposal of the lottery ticket hypothesis, researchers have focused either on pruning at initialization or after …
View article: Dynamics of Moral Behavior in Heterogeneous Populations of Learning Agents
Dynamics of Moral Behavior in Heterogeneous Populations of Learning Agents Open
Growing concerns about safety and alignment of AI systems highlight the importance of embedding moral capabilities in artificial agents: a promising solution is the use of learning from experience, i.e., Reinforcement Learning. In multi-ag…
View article: Moral Alignment for LLM Agents
Moral Alignment for LLM Agents Open
Decision-making agents based on pre-trained Large Language Models (LLMs) are increasingly being deployed across various domains of human activity. While their applications are currently rather specialized, several research efforts are unde…
View article: Partial Information Decomposition for Data Interpretability and Feature Selection
Partial Information Decomposition for Data Interpretability and Feature Selection Open
In this paper, we introduce Partial Information Decomposition of Features (PIDF), a new paradigm for simultaneous data interpretability and feature selection. Contrary to traditional methods that assign a single importance value, our appro…
View article: Large Language Models are Effective Priors for Causal Graph Discovery
Large Language Models are Effective Priors for Causal Graph Discovery Open
Causal structure discovery from observations can be improved by integrating background knowledge provided by an expert to reduce the hypothesis space. Recently, Large Language Models (LLMs) have begun to be considered as sources of prior i…
View article: Graph Reinforcement Learning for Combinatorial Optimization: A Survey and Unifying Perspective
Graph Reinforcement Learning for Combinatorial Optimization: A Survey and Unifying Perspective Open
Graphs are a natural representation for systems based on relations between connected entities. Combinatorial optimization problems, which arise when considering an objective function related to a process of interest on discrete structures,…
View article: Dynamics of Moral Behavior in Heterogeneous Populations of Learning Agents
Dynamics of Moral Behavior in Heterogeneous Populations of Learning Agents Open
Growing concerns about safety and alignment of AI systems highlight the importance of embedding moral capabilities in artificial agents: a promising solution is the use of learning from experience, i.e., Reinforcement Learning. In multi-ag…
View article: Information-Theoretic State Variable Selection for Reinforcement Learning
Information-Theoretic State Variable Selection for Reinforcement Learning Open
Identifying the most suitable variables to represent the state is a fundamental challenge in Reinforcement Learning (RL). These variables must efficiently capture the information necessary for making optimal decisions. In order to address …
View article: Hybrid Approaches for Moral Value Alignment in AI Agents: a Manifesto
Hybrid Approaches for Moral Value Alignment in AI Agents: a Manifesto Open
Increasing interest in ensuring the safety of next-generation Artificial Intelligence (AI) systems calls for novel approaches to embedding morality into autonomous agents. This goal differs qualitatively from traditional task-specific AI m…
View article: Identifying vulnerabilities of industrial control systems using evolutionary multiobjective optimisation
Identifying vulnerabilities of industrial control systems using evolutionary multiobjective optimisation Open
In this paper, we propose a novel methodology to assist in identifying vulnerabilities in real-world complex heterogeneous industrial control systems (ICS) using two Evolutionary Multiobjective Optimisation (EMO) algorithms, NSGA-II and SP…
View article: Tree Search in DAG Space with Model-based Reinforcement Learning for Causal Discovery
Tree Search in DAG Space with Model-based Reinforcement Learning for Causal Discovery Open
Identifying causal structure is central to many fields ranging from strategic decision-making to biology and economics. In this work, we propose CD-UCT, a model-based reinforcement learning method for causal discovery based on tree search …
View article: Modeling Moral Choices in Social Dilemmas with Multi-Agent Reinforcement Learning
Modeling Moral Choices in Social Dilemmas with Multi-Agent Reinforcement Learning Open
Practical uses of Artificial Intelligence (AI) in the real world have demonstrated the importance of embedding moral choices into intelligent agents. They have also highlighted that defining top-down ethical constraints on AI according to …
View article: Using behavioral studies to adapt management decisions and reduce negative interactions between humans and baboons in Cape Town, South Africa
Using behavioral studies to adapt management decisions and reduce negative interactions between humans and baboons in Cape Town, South Africa Open
Understanding the behavioral ecology of wildlife that experiences negative interactions with humans and the outcome of any wildlife management intervention is essential. In the Cape Peninsula, South Africa, chacma baboons ( Papio ursinus )…
View article: Machine Learning-based Intrusion Detection Systems: Deployment Guidelines for Industry
Machine Learning-based Intrusion Detection Systems: Deployment Guidelines for Industry Open
This report is meant to serve as a guide for operators, managers of industrial control systems or those responsible for making decisions related to designing, installing, purchasing, or maintaining the performance of Machine Learning-based…
View article: Modeling Moral Choices in Social Dilemmas with Multi-Agent Reinforcement Learning
Modeling Moral Choices in Social Dilemmas with Multi-Agent Reinforcement Learning Open
Practical uses of Artificial Intelligence (AI) in the real world have demonstrated the importance of embedding moral choices into intelligent agents. They have also highlighted that defining top-down ethical constraints on AI according to …
View article: Drawing on the Success of Developing a Safety Culture to Improve the Security Culture in Companies That Use Operational Technology
Drawing on the Success of Developing a Safety Culture to Improve the Security Culture in Companies That Use Operational Technology Open
Companies using operational technology (OT), including critical infrastructure ones, are increasingly becoming more digitalized.This digitalization, however, has led to an extended attack surface, making cybersecurity a necessity.One appro…
View article: rw_network_data.zip from Planning spatial networks with Monte Carlo tree search
rw_network_data.zip from Planning spatial networks with Monte Carlo tree search Open
Real-world network data (Internet dataset).
View article: Planning spatial networks with Monte Carlo tree search
Planning spatial networks with Monte Carlo tree search Open
We tackle the problem of goal-directed graph construction: given a starting graph, finding a set of edges whose addition maximally improves a global objective function. This problem emerges in many transportation and infrastructure network…
View article: source_code.zip from Planning spatial networks with Monte Carlo tree search
source_code.zip from Planning spatial networks with Monte Carlo tree search Open
Source code that enables reproducing the results reported in the paper.
View article: Graph Neural Modeling of Network Flows
Graph Neural Modeling of Network Flows Open
Network flow problems, which involve distributing traffic such that the underlying infrastructure is used effectively, are ubiquitous in transportation and logistics. Among them, the general Multi-Commodity Network Flow (MCNF) problem conc…