Alessandra Russo

Daniel Cunnington, Mark Law, Jorge Lobo et al.

One of the ultimate goals of Artificial Intelligence is to assist humans in complex decision making. A promising direction for achieving this goal is Neuro-Symbolic AI, which aims to combine the interpretability of symbolic techniques with the ability of deep learning to learn from raw data. However, most current approaches require manually engineered symbolic knowledge, and where end-to-end training is considered, such approaches are either restricted to learning definite programs, or are restricted to training binary neural networks. In this paper, we introduce Neuro-Symbolic Inductive Learner (NSIL), an approach that trains a general neural network to extract latent concepts from raw data, whilst learning symbolic knowledge that maps latent concepts to target labels. The novelty of our approach is a method for biasing the learning of symbolic knowledge, based on the in-training performance of both neural and symbolic components. We evaluate NSIL on three problem domains of different complexity, including an NP-complete problem. Our results demonstrate that NSIL learns expressive knowledge, solves computationally complex problems, and achieves state-of-the-art performance in terms of accuracy and data efficiency. Code and technical appendix: https://github.com/DanCunnington/NSIL

Jeya Vikranth Jeyakumar, Luke Dickens, Luis Garcia et al.

Recent efforts in interpretable deep learning models have shown that concept-based explanation methods achieve competitive accuracy with standard end-to-end models and enable reasoning and intervention about extracted high-level visual concepts from images, e.g., identifying the wing color and beak length for bird-species classification. However, these concept bottleneck models rely on a necessary and sufficient set of predefined concepts-which is intractable for complex tasks such as video classification. For complex tasks, the labels and the relationship between visual elements span many frames, e.g., identifying a bird flying or catching prey-necessitating concepts with various levels of abstraction. To this end, we present CoDEx, an automatic Concept Discovery and Extraction module that rigorously composes a necessary and sufficient set of concept abstractions for concept-based video classification. CoDEx identifies a rich set of complex concept abstractions from natural language explanations of videos-obviating the need to predefine the amorphous set of concepts. To demonstrate our method's viability, we construct two new public datasets that combine existing complex video classification datasets with short, crowd-sourced natural language explanations for their labels. Our method elicits inherent complex concept abstractions in natural language to generalize concept-bottleneck methods to complex tasks.

Julien Amblard, Niklas Groll, Matthew Tait et al.

Over the past decade, Artificial Intelligence has significantly advanced, mostly driven by large-scale neural approaches. However, in the chemical process industry, where safety is critical, these methods are often unsuitable due to their brittleness, and lack of explainability and interpretability. Furthermore, open-source real-world datasets containing historical failures are scarce in this domain. In this paper, we investigate an approach for predicting failures in chemical processes using symbolic machine learning and conduct a feasibility study in the context of an ethylene oxidation process. Our method builds on a state-of-the-art symbolic machine learning system capable of learning predictive models in the form of probabilistic rules from context-dependent noisy examples. This system is a general-purpose symbolic learner, which makes our approach independent of any specific chemical process. To address the lack of real-world failure data, we conduct our feasibility study leveraging data generated from a chemical process simulator. Experimental results show that symbolic machine learning can outperform baseline methods such as random forest and multilayer perceptron, while preserving interpretability through the generation of compact, rule-based predictive models. Finally, we explain how such learned rule-based models could be integrated into agents to assist chemical plant operators in decision-making during potential failures.

Daniel Furelos-Blanco, Mark Law, Anders Jonsson et al.

Reward machines (RMs) are a recent formalism for representing the reward function of a reinforcement learning task through a finite-state machine whose edges encode subgoals of the task using high-level events. The structure of RMs enables the decomposition of a task into simpler and independently solvable subtasks that help tackle long-horizon and/or sparse reward tasks. We propose a formalism for further abstracting the subtask structure by endowing an RM with the ability to call other RMs, thus composing a hierarchy of RMs (HRM). We exploit HRMs by treating each call to an RM as an independently solvable subtask using the options framework, and describe a curriculum-based method to learn HRMs from traces observed by the agent. Our experiments reveal that exploiting a handcrafted HRM leads to faster convergence than with a flat HRM, and that learning an HRM is feasible in cases where its equivalent flat representation is not.

Alex F. Spies, Alessandra Russo, Murray Shanahan

We investigate the composability of soft-rules learned by relational neural architectures when operating over object-centric (slot-based) representations, under a variety of sparsity-inducing constraints. We find that increasing sparsity, especially on features, improves the performance of some models and leads to simpler relations. Additionally, we observe that object-centric representations can be detrimental when not all objects are fully captured; a failure mode to which CNNs are less prone. These findings demonstrate the trade-offs between interpretability and performance, even for models designed to tackle relational tasks.

Leo Ardon, Daniel Furelos-Blanco, Alessandra Russo

This paper presents a novel approach to Multi-Agent Reinforcement Learning (MARL) that combines cooperative task decomposition with the learning of reward machines (RMs) encoding the structure of the sub-tasks. The proposed method helps deal with the non-Markovian nature of the rewards in partially observable environments and improves the interpretability of the learnt policies required to complete the cooperative task. The RMs associated with each sub-task are learnt in a decentralised manner and then used to guide the behaviour of each agent. By doing so, the complexity of a cooperative multi-agent problem is reduced, allowing for more effective learning. The results suggest that our approach is a promising direction for future research in MARL, especially in complex environments with large state spaces and multiple agents.

Charles Pert, Dalal Alrajeh, Alessandra Russo

Length generalization remains a persistent challenge for neural networks: recurrent models tend to suffer from positional biases, while transformers are constrained by fixed computational depth. Regular languages provide a frequently used testbed for evaluating length generalization, as label prediction can be checked for any sequence length. We propose MLP-LDRU, a type of Log-Depth Recurrent Unit, which captures a class of associativity-biased operators designed to approximate recurrence through parallel reduction. We evaluate MLP-LDRU on 21 regular-language tasks, consisting of standard benchmarks and new prefix languages, where it achieves 100% out-of-distribution accuracy on 18 tasks and at least 99.9% on the remaining 3 when increasing max training length, outperforming comparable recurrent and attention-based models. We further evaluate MLP-LDRU beyond regular languages on ListOps and NLP classification benchmarks, where it performs competitively.

Enrico Pontelli, Stefania Costantini, Carmine Dodaro et al.

This volume contains the Technical Communications presented at the 39th International Conference on Logic Programming (ICLP 2023), held at Imperial College London, UK from July 9 to July 15, 2023. Technical Communications included here concern the Main Track, the Doctoral Consortium, the Application and Systems/Demo track, the Recently Published Research Track, the Birds-of-a-Feather track, the Thematic Tracks on Logic Programming and Machine Learning, and Logic Programming and Explainability, Ethics, and Trustworthiness.

Mohammad Albinhassan, Pranava Madhyastha, Mark Law et al.

Controlling the output of Large Language Models (LLMs) through context-sensitive constraints has emerged as a promising approach to overcome the limitations of Context-Free Grammars (CFGs) in guaranteeing generation validity. However, such constraints typically require manual specification -- a significant barrier demanding specialized expertise. We introduce a framework that automatically learns context-sensitive constraints from LLM interactions through a two-phase process: syntactic exploration to gather diverse outputs for constraint learning, followed by constraint exploitation to enforce these learned rules during generation. Experiments demonstrate that our method enables even small LLMs (1B parameters) to learn and generate with perfect constraint adherence, outperforming larger counterparts and state-of-the-art reasoning models. This work represents the first integration of context-sensitive grammar learning with LLM generation, eliminating manual specification while maintaining generation validity.

Zlatina Mileva, Antonis Bikakis, Fabio Aurelio D'Asaro et al.

Argumentation is a very active research field of Artificial Intelligence concerned with the representation and evaluation of arguments used in dialogues between humans and/or artificial agents. Acceptability semantics of formal argumentation systems define the criteria for the acceptance or rejection of arguments. Several software systems, known as argumentation solvers, have been developed to compute the accepted/rejected arguments using such criteria. These include systems that learn to identify the accepted arguments using non-interpretable methods. In this paper we present a novel framework, which uses an Inductive Logic Programming approach to learn the acceptability semantics for several abstract and structured argumentation frameworks in an interpretable way. Through an empirical evaluation we show that our framework outperforms existing argumentation solvers, thus opening up new future research directions in the area of formal argumentation and human-machine dialogues.

Hadeel Al-Negheimish, Pranava Madhyastha, Alessandra Russo

Large pre-trained language models such as BERT have been widely used as a framework for natural language understanding (NLU) tasks. However, recent findings have revealed that pre-trained language models are insensitive to word order. The performance on NLU tasks remains unchanged even after randomly permuting the word of a sentence, where crucial syntactic information is destroyed. To help preserve the importance of word order, we propose a simple approach called Forced Invalidation (FI): forcing the model to identify permuted sequences as invalid samples. We perform an extensive evaluation of our approach on various English NLU and QA based tasks over BERT-based and attention-based models over word embeddings. Our experiments demonstrate that Forced Invalidation significantly improves the sensitivity of the models to word order.

Thomas Andrews, Mark Law, Sara Ahmadi-Abhari et al.

We introduce LearnAD, a neuro-symbolic method for predicting Alzheimer's disease from brain magnetic resonance imaging data, learning fully interpretable rules. LearnAD applies statistical models, Decision Trees, Random Forests, or GNNs to identify relevant brain connections, and then employs FastLAS to learn global rules. Our best instance outperforms Decision Trees, matches Support Vector Machine accuracy, and performs only slightly below Random Forests and GNNs trained on all features, all while remaining fully interpretable. Ablation studies show that our neuro-symbolic approach improves interpretability with comparable performance to pure statistical models. LearnAD demonstrates how symbolic learning can deepen our understanding of GNN behaviour in clinical neuroscience.

Roko Parac, Lorenzo Nodari, Leo Ardon et al.

This paper presents PROB-IRM, an approach that learns robust reward machines (RMs) for reinforcement learning (RL) agents from noisy execution traces. The key aspect of RM-driven RL is the exploitation of a finite-state machine that decomposes the agent's task into different subtasks. PROB-IRM uses a state-of-the-art inductive logic programming framework robust to noisy examples to learn RMs from noisy traces using the Bayesian posterior degree of beliefs, thus ensuring robustness against inconsistencies. Pivotal for the results is the interleaving between RM learning and policy learning: a new RM is learned whenever the RL agent generates a trace that is believed not to be accepted by the current RM. To speed up the training of the RL agent, PROB-IRM employs a probabilistic formulation of reward shaping that uses the posterior Bayesian beliefs derived from the traces. Our experimental analysis shows that PROB-IRM can learn (potentially imperfect) RMs from noisy traces and exploit them to train an RL agent to solve its tasks successfully. Despite the complexity of learning the RM from noisy traces, agents trained with PROB-IRM perform comparably to agents provided with handcrafted RMs.

Robson W. S. Pessoa, Julien Amblard, Alessandra Russo et al.

Anomaly detection in batch processes is hindered by transient dynamics, scarce fault labels, and reliance on single-modality sensor data. This work introduces UTOPYA (Unified Temporal Observation for Physics-Informed Anomaly Detection and Time-Series Prediction), a 15.2M-parameter multimodal framework that jointly addresses anomaly detection, time-series prediction, and phase classification in batch distillation by fusing eight data modalities through Feature-wise Linear Modulation (FiLM) conditioned cross-modal attention and gated fusion. A physics-informed regularisation scheme introduced in this work enforces temporal smoothness and thermodynamic monotonicity, while curriculum learning introduces training samples in order of physical difficulty. On the 119-experiment multimodal batch distillation dataset of Arweiler et al. (2026), UTOPYA achieves a window-level test AUROC of 0.832 and 0.874 under multi-signal experiment-level scoring, substantially outperforming four external baselines (PCA, autoencoder, Isolation Forest, and LSTM autoencoder) evaluated under identical conditions (+0.147 window-level AUROC over the best baseline). A multimodal ablation over 15~architectural configurations shows that static context via FiLM conditioning is the key enabler, lifting experiment-level multi-signal AUROC by +0.145 over the unimodal baseline (0.729 to 0.874). Separately, a training ablation across 14 design choices reveals that several widely-adopted techniques, including instance normalisation, Mixup, ensembling, test-time augmentation, and stochastic weight averaging, fail to improve or actively degrade generalisation in this data-scarce setting. These negative results expose a fundamental tension between smoothing-based regularisation and anomaly detection, providing practical guidance for multimodal process monitoring deployment.

Rakhilya Lee Mekhtieva, Brandon Forbes, Dalal Alrajeh et al.

Clinical decision-making is a fundamental stage in delivering appropriate care to patients. In recent years several decision-making systems designed to aid the clinician in this process have been developed. However, technical solutions currently in use are based on simple regression models and are only able to take into account simple pre-defined multiple-choice features, such as patient age, pre-existing conditions, smoker status, etc. One particular source of patient data, that available decision-making systems are incapable of processing is the collection of patient consultation GP notes. These contain crucial signs and symptoms - the information used by clinicians in order to make a final decision and direct the patient to the appropriate care. Extracting information from GP notes is a technically challenging problem, as they tend to include abbreviations, typos, and incomplete sentences. This paper addresses this open challenge. We present a framework that performs knowledge graph construction from raw GP medical notes written during or after patient consultations. By relying on support phrases mined from the SNOMED ontology, as well as predefined supported facts from values used in the RECAP (REmote COVID-19 Assessment in Primary Care) patient risk prediction tool, our graph generative framework is able to extract structured knowledge graphs from the highly unstructured and inconsistent format that consultation notes are written in. Our knowledge graphs include information about existing patient symptoms, their duration, and their severity. We apply our framework to consultation notes of COVID-19 patients in the UK COVID-19 Clinical Assesment Servcie (CCAS) patient dataset. We provide a quantitative evaluation of the performance of our framework, demonstrating that our approach has better accuracy than traditional NLP methods when answering questions about patients.

Lachlan McPheat, Navdeep Kaur, Robert Blackwell et al.

We introduce DecompSR, decomposed spatial reasoning, a large benchmark dataset (over 5m datapoints) and generation framework designed to analyse compositional spatial reasoning ability. The generation of DecompSR allows users to independently vary several aspects of compositionality, namely: productivity (reasoning depth), substitutivity (entity and linguistic variability), overgeneralisation (input order, distractors) and systematicity (novel linguistic elements). DecompSR is built procedurally in a manner which makes it is correct by construction, which is independently verified using a symbolic solver to guarantee the correctness of the dataset. DecompSR is comprehensively benchmarked across a host of Large Language Models (LLMs) where we show that LLMs struggle with productive and systematic generalisation in spatial reasoning tasks whereas they are more robust to linguistic variation. DecompSR provides a provably correct and rigorous benchmarking dataset with a novel ability to independently vary the degrees of several key aspects of compositionality, allowing for robust and fine-grained probing of the compositional reasoning abilities of LLMs.

Zihang Peng, Daria Stepanova, Vinh Thinh Ho et al.

Advances in information extraction have enabled the automatic construction of large knowledge graphs (e.g., Yago, Wikidata or Google KG), which are widely used in many applications like semantic search or data analytics. However, due to their semi-automatic construction, KGs are often incomplete. Rule learning methods, concerned with the extraction of frequent patterns from KGs and casting them into rules, can be applied to predict potentially missing facts. A crucial step in this process is rule ranking. Ranking of rules is especially challenging over highly incomplete or biased KGs (e.g., KGs predominantly storing facts about famous people), as in this case biased rules might fit the data best and be ranked at the top based on standard statistical metrics like rule confidence. To address this issue, prior works proposed to rank rules not only relying on the original KG but also facts predicted by a KG embedding model. At the same time, with the recent rise of Language Models (LMs), several works have claimed that LMs can be used as alternative means for KG completion. In this work, our goal is to verify to which extent the exploitation of LMs is helpful for improving the quality of rule learning systems.

Kexin Gu Baugh, Nuri Cingillioglu, Alessandra Russo

Neuro-symbolic rule learning has attracted lots of attention as it offers better interpretability than pure neural models and scales better than symbolic rule learning. A recent approach named pix2rule proposes a neural Disjunctive Normal Form (neural DNF) module to learn symbolic rules with feed-forward layers. Although proved to be effective in synthetic binary classification, pix2rule has not been applied to more challenging tasks such as multi-label and multi-class classifications over real-world data. In this paper, we address this limitation by extending the neural DNF module to (i) support rule learning in real-world multi-class and multi-label classification tasks, (ii) enforce the symbolic property of mutual exclusivity (i.e. predicting exactly one class) in multi-class classification, and (iii) explore its scalability over large inputs and outputs. We train a vanilla neural DNF model similar to pix2rule's neural DNF module for multi-label classification, and we propose a novel extended model called neural DNF-EO (Exactly One) which enforces mutual exclusivity in multi-class classification. We evaluate the classification performance, scalability and interpretability of our neural DNF-based models, and compare them against pure neural models and a state-of-the-art symbolic rule learner named FastLAS. We demonstrate that our neural DNF-based models perform similarly to neural networks, but provide better interpretability by enabling the extraction of logical rules. Our models also scale well when the rule search space grows in size, in contrast to FastLAS, which fails to learn in multi-class classification tasks with 200 classes and in all multi-label settings.

Christos Ziakas, Alessandra Russo, Avishek Joey Bose

Generative policies based on expressive model classes, such as diffusion and flow matching, are well-suited to complex control problems with highly multimodal action distributions. Their expressivity, however, comes at a significant inference cost: generating each action typically requires simulating many steps of the generative process, compounding latency across sequential decision-making rollouts. We introduce flow map policies, a novel class of generative policies designed for fast action generation by learning to take arbitrary-size jumps including one-step jumps-across the generative dynamics of existing flow-based policies. We instantiate flow map policies for offline-to-online reinforcement learning (RL) and formulate online adaptation as a trust-region optimization problem that improves the critic's Q-value while remaining close to the offline policy. We theoretically derive FLOW MAP Q-GUIDANCE (FMQ), a principled closed-form learning target that is optimal for adapting offline flow map policies under a critic-guided trust-region constraint. We further introduce Q-GUIDED BEAM SEARCH (QGBS), a stochastic flow-map sampler that combines renoising with beam search to enable iterative inference-time refinement. Across 12 challenging robotic manipulation and locomotion tasks from OGBench and RoboMimic, FMQ achieves state-of-the-art performance in offline-to-online RL, outperforming the previous one-step policy MVP by a relative improvement of 21.3% on the average success rate.

Theo Charalambous, Yaniv Aspis, Alessandra Russo

Symbolic rule learners generate interpretable solutions, however they require the input to be encoded symbolically. Neuro-symbolic approaches overcome this issue by mapping raw data to latent symbolic concepts using a neural network. Training the neural and symbolic components jointly is difficult, due to slow and unstable learning, hence many existing systems rely on hand-engineered rules to train the network. We introduce NeuralFastLAS, a scalable and fast end-to-end approach that trains a neural network jointly with a symbolic learner. For a given task, NeuralFastLAS computes a relevant set of rules, proved to contain an optimal symbolic solution, trains a neural network using these rules, and finally finds an optimal symbolic solution to the task while taking network predictions into account. A key novelty of our approach is learning a posterior distribution on rules while training the neural network to improve stability during training. We provide theoretical results for a sufficient condition on network training to guarantee correctness of the final solution. Experimental results demonstrate that NeuralFastLAS is able to achieve state-of-the-art accuracy in arithmetic and logical tasks, with a training time that is up to two orders of magnitude faster than other jointly trained neuro-symbolic methods.

Kexin Gu Baugh, Vincent Perreault, Matthew Baugh et al.

Neural Disjunctive Normal Form (DNF) based models are powerful and interpretable approaches to neuro-symbolic learning and have shown promising results in classification and reinforcement learning settings without prior knowledge of the tasks. However, their performance is degraded by the thresholding of the post-training symbolic translation process. We show here that part of the performance degradation during translation is due to its failure to disentangle the learned knowledge represented in the form of the networks' weights. We address this issue by proposing a new disentanglement method; by splitting nodes that encode nested rules into smaller independent nodes, we are able to better preserve the models' performance. Through experiments on binary, multiclass, and multilabel classification tasks (including those requiring predicate invention), we demonstrate that our disentanglement method provides compact and interpretable logical representations for the neural DNF-based models, with performance closer to that of their pre-translation counterparts. Our code is available at https://github.com/kittykg/disentangling-ndnf-classification.

Amrit Diggavi Seshadri, Alessandra Russo

In this work, following the intuition that adverbs describing scene-sequences are best identified by reasoning over high-level concepts of object-behavior, we propose the design of a new framework that reasons over object-behaviours extracted from raw-video-clips to recognize the clip's corresponding adverb-types. Importantly, while previous works for general scene adverb-recognition assume knowledge of the clips underlying action-types, our method is directly applicable in the more general problem setting where the action-type of a video-clip is unknown. Specifically, we propose a novel pipeline that extracts human-interpretable object-behaviour-facts from raw video clips and propose novel symbolic and transformer based reasoning methods that operate over these extracted facts to identify adverb-types. Experiment results demonstrate that our proposed methods perform favourably against the previous state-of-the-art. Additionally, to support efforts in symbolic video-processing, we release two new datasets of object-behaviour-facts extracted from raw video clips - the MSR-VTT-ASP and ActivityNet-ASP datasets.

Daniel Cunnington, Mark Law, Jorge Lobo et al.

Neuro-Symbolic AI (NeSy) holds promise to ensure the safe deployment of AI systems, as interpretable symbolic techniques provide formal behaviour guarantees. The challenge is how to effectively integrate neural and symbolic computation, to enable learning and reasoning from raw data. Existing pipelines that train the neural and symbolic components sequentially require extensive labelling, whereas end-to-end approaches are limited in terms of scalability, due to the combinatorial explosion in the symbol grounding problem. In this paper, we leverage the implicit knowledge within foundation models to enhance the performance in NeSy tasks, whilst reducing the amount of data labelling and manual engineering. We introduce a new architecture, called NeSyGPT, which fine-tunes a vision-language foundation model to extract symbolic features from raw data, before learning a highly expressive answer set program to solve a downstream task. Our comprehensive evaluation demonstrates that NeSyGPT has superior accuracy over various baselines, and can scale to complex NeSy tasks. Finally, we highlight the effective use of a large language model to generate the programmatic interface between the neural and symbolic components, significantly reducing the amount of manual engineering required.

Alex F. Spies, William Edwards, Michael I. Ivanitskiy et al.

Recent studies in interpretability have explored the inner workings of transformer models trained on tasks across various domains, often discovering that these networks naturally develop highly structured representations. When such representations comprehensively reflect the task domain's structure, they are commonly referred to as "World Models" (WMs). In this work, we identify WMs in transformers trained on maze-solving tasks. By using Sparse Autoencoders (SAEs) and analyzing attention patterns, we examine the construction of WMs and demonstrate consistency between SAE feature-based and circuit-based analyses. By subsequently intervening on isolated features to confirm their causal role, we find that it is easier to activate features than to suppress them. Furthermore, we find that models can reason about mazes involving more simultaneously active features than they encountered during training; however, when these same mazes (with greater numbers of connections) are provided to models via input tokens instead, the models fail. Finally, we demonstrate that positional encoding schemes appear to influence how World Models are structured within the model's residual stream.

Christos Ziakas, Amir Bar, Alessandra Russo

Large-scale video generative models have shown emerging capabilities as zero-shot visual planners, yet video-generated plans often violate temporal consistency and physical constraints, leading to failures when mapped to executable actions. To address this, we propose Grounding Video Plans with World Models (GVP-WM), a planning method that grounds video-generated plans into feasible action sequences using a learned action-conditioned world model. At test-time, GVP-WM first generates a video plan from initial and goal observations, then projects the video guidance onto the manifold of dynamically feasible latent trajectories via video-guided latent collocation. In particular, we formulate grounding as a goal-conditioned latent-space trajectory optimization problem that jointly optimizes latent states and actions under world-model dynamics, while preserving semantic alignment with the video-generated plan. Empirically, GVP-WM recovers feasible long-horizon plans from zero-shot image-to-video-generated and motion-blurred videos that violate physical constraints, across navigation and manipulation simulation tasks.

Kexin Gu Baugh, Luke Dickens, Alessandra Russo

Although deep reinforcement learning has been shown to be effective, the model's black-box nature presents barriers to direct policy interpretation. To address this problem, we propose a neuro-symbolic approach called neural DNF-MT for end-to-end policy learning. The differentiable nature of the neural DNF-MT model enables the use of deep actor-critic algorithms for training. At the same time, its architecture is designed so that trained models can be directly translated into interpretable policies expressed as standard (bivalent or probabilistic) logic programs. Moreover, additional layers can be included to extract abstract features from complex observations, acting as a form of predicate invention. The logic representations are highly interpretable, and we show how the bivalent representations of deterministic policies can be edited and incorporated back into a neural model, facilitating manual intervention and adaptation of learned policies. We evaluate our approach on a range of tasks requiring learning deterministic or stochastic behaviours from various forms of observations. Our empirical results show that our neural DNF-MT model performs at the level of competing black-box methods whilst providing interpretable policies.

Leo Ardon, Daniel Furelos-Blanco, Roko Parac et al.

Reward machines (RMs) are an effective approach for addressing non-Markovian rewards in reinforcement learning (RL) through finite-state machines. Traditional RMs, which label edges with propositional logic formulae, inherit the limited expressivity of propositional logic. This limitation hinders the learnability and transferability of RMs since complex tasks will require numerous states and edges. To overcome these challenges, we propose First-Order Reward Machines ($\texttt{FORM}$s), which use first-order logic to label edges, resulting in more compact and transferable RMs. We introduce a novel method for $\textbf{learning}$ $\texttt{FORM}$s and a multi-agent formulation for $\textbf{exploiting}$ them and facilitate their transferability, where multiple agents collaboratively learn policies for a shared $\texttt{FORM}$. Our experimental results demonstrate the scalability of $\texttt{FORM}$s with respect to traditional RMs. Specifically, we show that $\texttt{FORM}$s can be effectively learnt for tasks where traditional RM learning approaches fail. We also show significant improvements in learning speed and task transferability thanks to the multi-agent learning framework and the abstraction provided by the first-order language.

Ruben Vereecken, Luke Dickens, Alessandra Russo

A key challenge in scaling up Reinforcement Learning is generalizing learned behaviour. Without the ability to carry forward acquired knowledge an agent is doomed to learn each task from scratch. In this paper we develop a new formalism for transfer by virtue of state abstraction. Based on task-independent, compact observations (outcomes) of the environment, we introduce Outcome-Predictive State Representations (OPSRs), agent-centered and task-independent abstractions that are made up of predictions of outcomes. We show formally and empirically that they have the potential for optimal but limited transfer, then overcome this trade-off by introducing OPSR-based skills, i.e. abstract actions (based on options) that can be reused between tasks as a result of state abstraction. In a series of empirical studies, we learn OPSR-based skills from demonstrations and show how they speed up learning considerably in entirely new and unseen tasks without any pre-processing. We believe that the framework introduced in this work is a promising step towards transfer in RL in general, and towards transfer through combining state and action abstraction specifically.

Navdeep Kaur, Lachlan McPheat, Alessandra Russo et al.

In this paper, we examine the use of Conformal Language Modelling (CLM) alongside Answer Set Programming (ASP) to enhance the performance of standard open-weight LLMs on complex multi-step reasoning tasks. Using the StepGame dataset, which requires spatial reasoning, we apply CLM to generate sets of ASP programs from an LLM, providing statistical guarantees on the correctness of the outputs. Experimental results show that CLM significantly outperforms baseline models that use standard sampling methods, achieving substantial accuracy improvements across different levels of reasoning complexity. Additionally, the LLM-as-Judge metric enhances CLM's performance, especially in assessing structurally and logically correct ASP outputs. However, calibrating CLM with diverse calibration sets did not improve generalizability for tasks requiring much longer reasoning steps, indicating limitations in handling more complex tasks.

Daniel Furelos-Blanco, Charles Pert, Frederik Kelbel et al.

Training general agents to follow complex instructions (tasks) in intricate environments (levels) remains a core challenge in reinforcement learning. Random sampling of task-level pairs often produces unsolvable combinations, highlighting the need to co-design tasks and levels. While unsupervised environment design (UED) has proven effective at automatically designing level curricula, prior work has only considered a fixed task. We present ATLAS (Aligning Tasks and Levels for Autocurricula of Specifications), a novel method that generates joint autocurricula over tasks and levels. Our approach builds upon UED to automatically produce solvable yet challenging task-level pairs for policy training. To evaluate ATLAS and drive progress in the field, we introduce an evaluation suite that models tasks as reward machines in Minigrid levels. Experiments demonstrate that ATLAS vastly outperforms random sampling approaches, particularly when sampling solvable pairs is unlikely. We further show that mutations leveraging the structure of both tasks and levels accelerate convergence to performant policies.

Christos Ziakas, Nicholas Loo, Nishita Jain et al.

Automated red-teaming has emerged as a scalable approach for auditing Large Language Models (LLMs) prior to deployment, yet existing approaches lack mechanisms to efficiently adapt to model-specific vulnerabilities at inference. We introduce Red-Bandit, a red-teaming framework that adapts online to identify and exploit model failure modes under distinct attack styles (e.g., manipulation, slang). Red-Bandit post-trains a set of parameter-efficient LoRA experts, each specialized for a particular attack style, using reinforcement learning that rewards the generation of unsafe prompts via a rule-based safety model. At inference, a multi-armed bandit policy dynamically selects among these attack-style experts based on the target model's response safety, balancing exploration and exploitation. Red-Bandit achieves state-of-the-art results on AdvBench under sufficient exploration (ASR@10), while producing more human-readable prompts (lower perplexity). Moreover, Red-Bandit's bandit policy serves as a diagnostic tool for uncovering model-specific vulnerabilities by indicating which attack styles most effectively elicit unsafe behaviors.

Manuel Borroto, Katie Gallagher, Antonio Ielo et al.

Large Language Models (LLMs) excel at understanding natural language but struggle with explicit commonsense reasoning. A recent trend of research suggests that the combination of LLM with robust symbolic reasoning systems can overcome this problem on story-based question answering tasks. In this setting, existing approaches typically depend on human expertise to manually craft the symbolic component. We argue, however, that this component can also be automatically learned from examples. In this work, we introduce LLM2LAS, a hybrid system that effectively combines the natural language understanding capabilities of LLMs, the rule induction power of the Learning from Answer Sets (LAS) system ILASP, and the formal reasoning strengths of Answer Set Programming (ASP). LLMs are used to extract semantic structures from text, which ILASP then transforms into interpretable logic rules. These rules allow an ASP solver to perform precise and consistent reasoning, enabling correct answers to previously unseen questions. Empirical results outline the strengths and weaknesses of our automatic approach for learning and reasoning in a story-based question answering benchmark.

Charles Pert, Dalal Alrajeh, Alessandra Russo

Büchi automata (BAs) recognize $ω$-regular languages defined by formal specifications like linear temporal logic (LTL) and are commonly used in the verification of reactive systems. However, BAs face scalability challenges when handling and manipulating complex system behaviors. As neural networks are increasingly used to address these scalability challenges in areas like model checking, investigating their ability to generalize beyond training data becomes necessary. This work presents the first study investigating whether recurrent neural networks (RNNs) can generalize to $ω$-regular languages derived from LTL formulas. We train RNNs on ultimately periodic $ω$-word sequences to replicate target BA behavior and evaluate how well they generalize to out-of-distribution sequences. Through experiments on LTL formulas corresponding to deterministic automata of varying structural complexity, from 3 to over 100 states, we show that RNNs achieve high accuracy on their target $ω$-regular languages when evaluated on sequences up to $8 \times$ longer than training examples, with $92.6\%$ of tasks achieving perfect or near-perfect generalization. These results establish the feasibility of neural approaches for learning complex $ω$-regular languages, suggesting their potential as components in neurosymbolic verification methods.

Alessandro Gaudenzi, Lorenzo Nodari, Lance Kaplan et al.

Advanced Persistent Threats (APTs) represent a significant challenge in cybersecurity due to their prolonged, multi-stage nature and the sophistication of their operators. Traditional detection systems typically focus on identifying malicious activity in binary terms (benign or malicious) without accounting for the progression of an attack. However, effective response strategies depend on accurate inference of the attack's current stage, as countermeasures must be tailored to whether an adversary is in the early reconnaissance phase or actively conducting exploitation or exfiltration. This work addresses the problem of attack stage inference under uncertainty, with a focus on robustness to out-of-distribution (OOD) inputs. We propose a classification approach based on Evidential Deep Learning (EDL), which models predictive uncertainty by outputting parameters of a Dirichlet distribution over possible stages. This allows the system not only to predict the most likely stage of an attack but also to indicate when it is uncertain or the input lies outside the training distribution. Preliminary experiments in a simulated environment demonstrate that the proposed model can accurately infer the stage of an attack with calibrated confidence while effectively detecting OOD inputs, which may indicate changes in the attackers' tactics. These results support the feasibility of deploying uncertainty-aware models for staged threat detection in dynamic and adversarial environments.

Christos Ziakas, Alessandra Russo

Vision-Language Models (VLMs) show promise as zero-shot goal-conditioned value functions, but their frozen pre-trained representations limit generalization and temporal reasoning. We introduce VITA, a zero-shot value function learning method that enhances both capabilities via test-time adaptation. At inference, a lightweight adaptation module is updated via a gradient step on a meta-learned self-supervised loss, such that each test-time update improves value estimation. By updating sequentially over a trajectory, VITA encodes history into its parameters, addressing the temporal reasoning limitations. To mitigate shortcut learning, we propose a dissimilarity-based sampling strategy that selects semantically diverse segments of the trajectory during training. In real-world robotic manipulation tasks, VITA generalizes from a single training environment to diverse out-of-distribution tasks, environments, and embodiments, outperforming the state-of-the-art zero-shot method using autoregressive VLMs. Furthermore, we demonstrate that VITA's zero-shot value estimates can be utilized for reward shaping in offline reinforcement learning, resulting in multi-task policies on the Meta-World benchmark that exceed the performance of those trained with the simulation's fuzzy-logic dense rewards.

Mohammad Albinhassan, Pranava Madhyastha, Alessandra Russo

Ensuring both syntactic and semantic correctness in Large Language Model (LLM) outputs remains a significant challenge, despite being critical for real-world deployment. In this paper, we introduce $\texttt{SEM-CTRL}$, a unified approach that enforces rich context-sensitive constraints and task- and instance-specific semantics directly on an LLM decoder. Our approach integrates token-level MCTS, which is guided by specific syntactic and semantic constraints. The constraints over the desired outputs are expressed using Answer Set Grammars -- a logic-based formalism that generalizes context-sensitive grammars while incorporating background knowledge to represent task-specific semantics. We show that our approach guarantees correct completions for any off-the-shelf LLM without the need for fine-tuning. We evaluate $\texttt{SEM-CTRL}$ on a range of tasks, including synthetic grammar synthesis, combinatorial reasoning, and planning. Our results demonstrate that $\texttt{SEM-CTRL}$ allows small pre-trained LLMs to efficiently outperform larger variants and state-of-the-art reasoning models (e.g., o1-preview) while simultaneously guaranteeing solution correctness.

Hadeel Al-Negheimish, Pranava Madhyastha, Alessandra Russo

Numerical reasoning based machine reading comprehension is a task that involves reading comprehension along with using arithmetic operations such as addition, subtraction, sorting, and counting. The DROP benchmark (Dua et al., 2019) is a recent dataset that has inspired the design of NLP models aimed at solving this task. The current standings of these models in the DROP leaderboard, over standard metrics, suggest that the models have achieved near-human performance. However, does this mean that these models have learned to reason? In this paper, we present a controlled study on some of the top-performing model architectures for the task of numerical reasoning. Our observations suggest that the standard metrics are incapable of measuring progress towards such tasks.

Daniel Cunnington, Mark Law, Alessandra Russo et al.

Logic-based machine learning aims to learn general, interpretable knowledge in a data-efficient manner. However, labelled data must be specified in a structured logical form. To address this limitation, we propose a neural-symbolic learning framework, called Feed-Forward Neural-Symbolic Learner (FFNSL), that integrates a logic-based machine learning system capable of learning from noisy examples, with neural networks, in order to learn interpretable knowledge from labelled unstructured data. We demonstrate the generality of FFNSL on four neural-symbolic classification problems, where different pre-trained neural network models and logic-based machine learning systems are integrated to learn interpretable knowledge from sequences of images. We evaluate the robustness of our framework by using images subject to distributional shifts, for which the pre-trained neural networks may predict incorrectly and with high confidence. We analyse the impact that these shifts have on the accuracy of the learned knowledge and run-time performance, comparing FFNSL to tree-based and pure neural approaches. Our experimental results show that FFNSL outperforms the baselines by learning more accurate and interpretable knowledge with fewer examples.

Nuri Cingillioglu, Alessandra Russo

Humans have the ability to seamlessly combine low-level visual input with high-level symbolic reasoning often in the form of recognising objects, learning relations between them and applying rules. Neuro-symbolic systems aim to bring a unifying approach to connectionist and logic-based principles for visual processing and abstract reasoning respectively. This paper presents a complete neuro-symbolic method for processing images into objects, learning relations and logical rules in an end-to-end fashion. The main contribution is a differentiable layer in a deep learning architecture from which symbolic relations and rules can be extracted by pruning and thresholding. We evaluate our model using two datasets: subgraph isomorphism task for symbolic rule learning and an image classification domain with compound relations for learning objects, relations and rules. We demonstrate that our model scales beyond state-of-the-art symbolic learners and outperforms deep relational neural network architectures.

Hadeel Al-Negheimish, Pranava Madhyastha, Alessandra Russo

Reasoning about information from multiple parts of a passage to derive an answer is an open challenge for reading-comprehension models. In this paper, we present an approach that reasons about complex questions by decomposing them to simpler subquestions that can take advantage of single-span extraction reading-comprehension models, and derives the final answer according to instructions in a predefined reasoning template. We focus on subtraction-based arithmetic questions and evaluate our approach on a subset of the DROP dataset. We show that our approach is competitive with the state-of-the-art while being interpretable and requires little supervision

Theophile Sautory, Nuri Cingillioglu, Alessandra Russo

The task of Video Question Answering (VideoQA) consists in answering natural language questions about a video and serves as a proxy to evaluate the performance of a model in scene sequence understanding. Most methods designed for VideoQA up-to-date are end-to-end deep learning architectures which struggle at complex temporal and causal reasoning and provide limited transparency in reasoning steps. We present the HySTER: a Hybrid Spatio-Temporal Event Reasoner to reason over physical events in videos. Our model leverages the strength of deep learning methods to extract information from video frames with the reasoning capabilities and explainability of symbolic artificial intelligence in an answer set programming framework. We define a method based on general temporal, causal and physics rules which can be transferred across tasks. We apply our model to the CLEVRER dataset and demonstrate state-of-the-art results in question answering accuracy. This work sets the foundations for the incorporation of inductive logic programming in the field of VideoQA.

Daniel Cunnington, Alessandra Russo, Mark Law et al.

Inductive Logic Programming (ILP) systems learn generalised, interpretable rules in a data-efficient manner utilising existing background knowledge. However, current ILP systems require training examples to be specified in a structured logical format. Neural networks learn from unstructured data, although their learned models may be difficult to interpret and are vulnerable to data perturbations at run-time. This paper introduces a hybrid neural-symbolic learning framework, called NSL, that learns interpretable rules from labelled unstructured data. NSL combines pre-trained neural networks for feature extraction with FastLAS, a state-of-the-art ILP system for rule learning under the answer set semantics. Features extracted by the neural components define the structured context of labelled examples and the confidence of the neural predictions determines the level of noise of the examples. Using the scoring function of FastLAS, NSL searches for short, interpretable rules that generalise over such noisy examples. We evaluate our framework on propositional and first-order classification tasks using the MNIST dataset as raw data. Specifically, we demonstrate that NSL is able to learn robust rules from perturbed MNIST data and achieve comparable or superior accuracy when compared to neural network and random forest baselines whilst being more general and interpretable.

Harald Strömfelt, Luke Dickens, Artur d'Avila Garcez et al.

We propose a new model for relational VAE semi-supervision capable of balancing disentanglement and low complexity modelling of relations with different symbolic properties. We compare the relative benefits of relation-decoder complexity and latent space structure on both inductive and transductive transfer learning. Our results depict a complex picture where enforcing structure on semi-supervised representations can greatly improve zero-shot transductive transfer, but may be less favourable or even impact negatively the capacity for inductive transfer.

Amani Abu Jabal, Elisa Bertino, Jorge Lobo et al.

Technology advances in areas such as sensors, IoT, and robotics, enable new collaborative applications (e.g., autonomous devices). A primary requirement for such collaborations is to have a secure system which enables information sharing and information flow protection. Policy-based management system is a key mechanism for secure selective sharing of protected resources. However, policies in each party of such a collaborative environment cannot be static as they have to adapt to different contexts and situations. One advantage of collaborative applications is that each party in the collaboration can take advantage of knowledge of the other parties for learning or enhancing its own policies. We refer to this learning mechanism as policy transfer. The design of a policy transfer framework has challenges, including policy conflicts and privacy issues. Policy conflicts typically arise because of differences in the obligations of the parties, whereas privacy issues result because of data sharing constraints for sensitive data. Hence, the policy transfer framework should be able to tackle such challenges by considering minimal sharing of data and support policy adaptation to address conflict. In the paper we propose a framework that aims at addressing such challenges. We introduce a formal definition of the policy transfer problem for attribute-based policies. We then introduce the transfer methodology that consists of three sequential steps. Finally we report experimental results.

Francesco Ricca, Alessandra Russo, Sergio Greco et al.

Since the first conference held in Marseille in 1982, ICLP has been the premier international event for presenting research in logic programming. Contributions are solicited in all areas of logic programming and related areas, including but not restricted to: - Foundations: Semantics, Formalisms, Answer-Set Programming, Non-monotonic Reasoning, Knowledge Representation. - Declarative Programming: Inference engines, Analysis, Type and mode inference, Partial evaluation, Abstract interpretation, Transformation, Validation, Verification, Debugging, Profiling, Testing, Logic-based domain-specific languages, constraint handling rules. - Related Paradigms and Synergies: Inductive and Co-inductive Logic Programming, Constraint Logic Programming, Interaction with SAT, SMT and CSP solvers, Logic programming techniques for type inference and theorem proving, Argumentation, Probabilistic Logic Programming, Relations to object-oriented and Functional programming, Description logics, Neural-Symbolic Machine Learning, Hybrid Deep Learning and Symbolic Reasoning. - Implementation: Concurrency and distribution, Objects, Coordination, Mobility, Virtual machines, Compilation, Higher Order, Type systems, Modules, Constraint handling rules, Meta-programming, Foreign interfaces, User interfaces. - Applications: Databases, Big Data, Data Integration and Federation, Software Engineering, Natural Language Processing, Web and Semantic Web, Agents, Artificial Intelligence, Bioinformatics, Education, Computational life sciences, Education, Cybersecurity, and Robotics.

Daniel Furelos-Blanco, Mark Law, Anders Jonsson et al.

In this paper we present ISA, an approach for learning and exploiting subgoals in episodic reinforcement learning (RL) tasks. ISA interleaves reinforcement learning with the induction of a subgoal automaton, an automaton whose edges are labeled by the task's subgoals expressed as propositional logic formulas over a set of high-level events. A subgoal automaton also consists of two special states: a state indicating the successful completion of the task, and a state indicating that the task has finished without succeeding. A state-of-the-art inductive logic programming system is used to learn a subgoal automaton that covers the traces of high-level events observed by the RL agent. When the currently exploited automaton does not correctly recognize a trace, the automaton learner induces a new automaton that covers that trace. The interleaving process guarantees the induction of automata with the minimum number of states, and applies a symmetry breaking mechanism to shrink the search space whilst remaining complete. We evaluate ISA in several gridworld and continuous state space problems using different RL algorithms that leverage the automaton structures. We provide an in-depth empirical analysis of the automaton learning performance in terms of the traces, the symmetry breaking and specific restrictions imposed on the final learnable automaton. For each class of RL problem, we show that the learned automata can be successfully exploited to learn policies that reach the goal, achieving an average reward comparable to the case where automata are not learned but handcrafted and given beforehand.

Daniel Pace, Alessandra Russo, Murray Shanahan

The i.i.d. assumption is a useful idealization that underpins many successful approaches to supervised machine learning. However, its violation can lead to models that learn to exploit spurious correlations in the training data, rendering them vulnerable to adversarial interventions, undermining their reliability, and limiting their practical application. To mitigate this problem, we present a method for learning multiple models, incorporating an objective that pressures each to learn a distinct way to solve the task. We propose a notion of diversity based on minimizing the conditional total correlation of final layer representations across models given the label, which we approximate using a variational estimator and minimize using adversarial training. To demonstrate our framework's ability to facilitate rapid adaptation to distribution shift, we train a number of simple classifiers from scratch on the frozen outputs of our models using a small amount of data from the shifted distribution. Under this evaluation protocol, our framework significantly outperforms a baseline trained using the empirical risk minimization principle.

Mark Law, Alessandra Russo, Krysia Broda

The goal of Inductive Logic Programming (ILP) is to learn a program that explains a set of examples in the context of some pre-existing background knowledge. Until recently, most research on ILP targeted learning Prolog programs. Our own ILASP system instead learns Answer Set Programs, including normal rules, choice rules and hard and weak constraints. Learning such expressive programs widens the applicability of ILP considerably; for example, enabling preference learning, learning common-sense knowledge, including defaults and exceptions, and learning non-deterministic theories. In this paper, we first give a general overview of ILASP's learning framework and its capabilities. This is followed by a comprehensive summary of the evolution of the ILASP system, presenting the strengths and weaknesses of each version, with a particular emphasis on scalability.

David Tuckey, Alessandra Russo, Krysia Broda

Explainability in AI is gaining attention in the computer science community in response to the increasing success of deep learning and the important need of justifying how such systems make predictions in life-critical applications. The focus of explainability in AI has predominantly been on trying to gain insights into how machine learning systems function by exploring relationships between input data and predicted outcomes or by extracting simpler interpretable models. Through literature surveys of philosophy and social science, authors have highlighted the sharp difference between these generated explanations and human-made explanations and claimed that current explanations in AI do not take into account the complexity of human interaction to allow for effective information passing to not-expert users. In this paper we instantiate the concept of structure of scientific explanation as the theoretical underpinning for a general framework in which explanations for AI systems can be implemented. This framework aims to provide the tools to build a "mental-model" of any AI system so that the interaction with the user can provide information on demand and be closer to the nature of human-made explanations. We illustrate how we can utilize this framework through two very different examples: an artificial neural network and a Prolog solver and we provide a possible implementation for both examples.

Daniel Furelos-Blanco, Mark Law, Alessandra Russo et al.

In this work we present ISA, a novel approach for learning and exploiting subgoals in reinforcement learning (RL). Our method relies on inducing an automaton whose transitions are subgoals expressed as propositional formulas over a set of observable events. A state-of-the-art inductive logic programming system is used to learn the automaton from observation traces perceived by the RL agent. The reinforcement learning and automaton learning processes are interleaved: a new refined automaton is learned whenever the RL agent generates a trace not recognized by the current automaton. We evaluate ISA in several gridworld problems and show that it performs similarly to a method for which automata are given in advance. We also show that the learned automata can be exploited to speed up convergence through reward shaping and transfer learning across multiple tasks. Finally, we analyze the running time and the number of traces that ISA needs to learn an automata, and the impact that the number of observable events has on the learner's performance.