Publications

[j1]

Neuro-Symbolic Predictive Process Monitoring

Axel Mezini, Elena Umili, Ivan Donadello, Fabrizio Maria Maggi, Matteo Mancanelli, Fabio Patrizi

Information Systems (IS), Volume 141, 2026.

[abstract] [paper] [bibtex]

This paper addresses the problem of suffix prediction in Business Process Management (BPM) by proposing a Neuro-Symbolic Predictive Process Monitoring (PPM) approach that integrates data-driven learning with temporal logic-based prior knowledge. While recent approaches leverage deep learning models for suffix prediction, they often fail to satisfy even basic logical constraints due to the absence of explicit integration of domain knowledge during training. We propose a novel method to incorporate Linear Temporal Logic over finite traces (LTLf) into the training process of autoregressive sequence predictors. Our approach introduces a differentiable logical loss function, defined using a soft approximation of LTLf semantics and the Gumbel-Softmax trick, which can be combined with standard predictive losses. This ensures the model learns to generate suffixes that are both accurate and logically consistent. Experimental evaluation on three real-world datasets shows that our method improves suffix prediction accuracy and compliance with temporal constraints. We also introduce two variants of the logic loss (local and global) and demonstrate their effectiveness under noisy and realistic settings. While developed in the context of BPM, our framework is applicable to any symbolic sequence generation task and contributes toward advancing Neuro-Symbolic AI.

@article{mezini2026neurosymbolic,
title = {Neuro-Symbolic Predictive Process Monitoring},
author = {Mezini, Axel and Umili, Elena and Donadello, Ivan and Maggi, Fabrizio Maria and Mancanelli, Matteo and Patrizi, Fabio},
journal = {Information Systems},
volume = {141},
pages = {102762},
year = {2026},
publisher = {Elsevier}
issn = {0306-4379}
doi = {https://doi.org/10.1016/j.is.2026.102762}
}

[c8]

Reactive Synthesis for Golog Specifications in the Propositional Situation Calculus

Giuseppe De Giacomo, Yves Lespérance, Matteo Mancanelli, Gianmarco Parretti

To appear in Proc. of the 22nd International Conference on Principles of Knowledge Representation and Reasoning (KR 2026), part of FLoC 2026, Lisbon, Portugal.

[abstract] [paper] [bibtex]

Golog programs over Situation Calculus action theories were introduced as a specification of desired agent behavior, very much like temporally extended goals in planning, but with a focus on procedural aspects typical of programs. In the words of the original paper: "Golog allows the programmer to strike a compromise between the often computationally infeasible classical planning task, in which a plan must be deduced entirely from scratch, and detailed programming, in which every little step must be specified." In this paper, we study temporal synthesis with Golog programs as specifications over nondeterministic propositional action theories. We show that Golog has the same expressive power as linear dynamic logics on finite traces (LDLf), namely that of regular languages or monadic second-order logic (MSO) over finite traces, while exhibiting a markedly lower synthesis complexity: synthesis can be performed by constructing a polynomial-size program graph and taking its cross-product with the domain, whereas LDLf synthesis requires building a deterministic automaton of worst-case doubly exponential size. This advantage is confirmed experimentally.

@inproceedings{de2026reactive,
title = {Reactive Synthesis for Golog Specifications in the Propositional Situation Calculus},
author = {De Giacomo, Giuseppe and Lesp{\'e}rance, Yves and Mancanelli, Matteo and Parretti, Gianmarco},
booktitle = {Proceedings of the 22nd International Conference on Principles of Knowledge Representation and Reasoning},
year = {2026}
}

[c7]

Neuro-Symbolic Injection of LTLf Constraints in Autoregressive Reinforcement Learning Policies

Ashkan Ansarifard, Matteo Mancanelli, Elena Umili, Fabio Patrizi

To appear in Joint Workshop on Statistics and Knowledge Integration for Logic, Learning, Ethical Decisions, and LLMs (SKILLED-LLMs 2026), co-located with KR 2026 and FLoC 2026, Lisbon, Portugal.

[abstract] [paper] [bibtex]

In this work we study offline reinforcement learning (RL) under temporally extended task constraints expressed in Linear Temporal Logic over finite traces (LTLf). Recently, transformer-based approaches such as Trajectory Transformers and Decision Transformers have been adopted to address RL as a sequence modeling problem. However, these methods optimize purely for reward and do not account for high-level temporal requirements. Here, we introduce a neurosymbolic framework that injects LTLf background knowledge into such transformer-based RL policies. Our approach compiles LTLf formulas into deterministic finite automata (DFAs) and integrates them into the learning process through a differentiable representation and a logic-based loss function. In particular, we derive differentiable satisfaction signals from DFA progression and use them as a regularization term during training. The resulting method is architecture-agnostic across different models. We evaluate the proposed framework on navigation environments with specification suites covering combinations of safety and reachability temporal properties. Experimental results show that incorporating background knowledge not only improves constraint satisfaction, but also maintains competitive return compared to vanilla baselines.

@inproceedings{ansarifard2026neuro,
title = {Neuro-Symbolic Injection of {LTLf} Constraints in Autoregressive Reinforcement Learning Policies},
author = {Ansarifard, Ashkan and Mancanelli, Matteo and Umili, Elena and Patrizi, Fabio},
booktitle = {Proceedings of the Joint Workshop on Statistics and Knowledge Integration for Logic, Learning, Ethical Decisions, and LLMs (SKILLED-LLMs 2026)},
year = {2026}
}

[c6]

An Incremental Method for Synthesizing Action Theory Abstractions

Bita Banihashemi, Yves Lespérance, Matteo Mancanelli

To appear in Workshop on Theory and Methods for Abstraction (THEMA 2026), co-located with KR 2026 and FLoC 2026, Lisbon, Portugal.

[abstract] [paper] [bibtex]

Abstraction is widely used in reasoning about action, for instance to facilitate planning or to provide explanations of agent behavior. In this paper, we present an incremental method to synthesize an abstract basic action theory (BAT) in the situation calculus that suppresses uninteresting details from a concrete BAT based on specifications provided by a knowledge engineer. The engineer can introduce abstract actions that are implemented by/mapped into Golog programs over the low-level theory and drop some low-level actions from the abstract BAT. She can also perform state abstraction by introducing high-level fluents that are mapped to low-level state formulas and drop some low-level fluents from the abstract BAT. The method allows the engineer to synthesize the abstract BAT step-by-step, providing the mapping along the way. The abstraction steps are validated to ensure that the resulting high-level BAT is guaranteed to be a sound and/or complete abstraction of the original low-level BAT. Abstraction steps that would fail to yield a sound/complete abstraction are rejected with an explanation for the failure, which guides the engineer towards a solution. Regression and SMT reasoning are used to automatically generate the axiomatization of the new abstract actions and fluents.

@inproceedings{banihashemi2026incremental,
title = {An Incremental Method for Synthesizing Action Theory Abstractions},
author = {Banihashemi, Bita and Lesp{\'e}rance, Yves and Mancanelli, Matteo},
booktitle = {Proceedings of the Workshop on Theory and Methods for Abstraction (THEMA 2026)},
year = {2026}
}

[c5]

Strategic Reasoning over Golog Programs in the Nondeterministic Situation Calculus

Giuseppe De Giacomo, Yves Lespérance, Matteo Mancanelli

Proc. of the 40th Annual AAAI Conference on Artificial Intelligence (AAAI 2026), Singapore, SG.

[abstract] [paper] [bibtex]

We investigate the problem of synthesizing strategies that guarantee the successful execution of a high-level nondeterministic agent program δ in Golog within a nondeterministic first-order basic action theory, considering the environment as adversarial. Our approach constructs a symbolic program graph that captures the control flow of δ independently of the domain, enabling strategy synthesis through the cross product of the program graph with the domain model. We formally relate graph-based transitions to standard Golog semantics and provide a synthesis procedure that is sound though incomplete (in general, the problem is undecidable, given that we have a first-order representation of the state). We also extend the framework to handle the case where the environment's possible behaviors are specified by a Golog program.

@inproceedings{de2026strategic,
title = {Strategic Reasoning over Golog Programs in the Nondeterministic Situation Calculus},
author = {De Giacomo, Giuseppe and Lesp{\'e}rance, Yves and Mancanelli, Matteo},
booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},
volume = {40},
number = {23},
pages = {19047--19054},
year = {2026}
}

[c4]

Strategic Reasoning over Golog Programs in the Nondeterministic Situation Calculus - Extended Abstract

Giuseppe De Giacomo, Yves Lespérance, Matteo Mancanelli

7th International Workshop on Artificial Intelligence and fOrmal VERification, Logic Automata, and sYnthesis (OVERLAY 2025), co-located with ECAI 2025, Bologna, Italy.

[abstract] [paper] [bibtex]

Automata-based synthesis has seen increasing interest in the last decade, mostly focused on declarative specifications. Here, we consider procedural programs that specify high-level agent behaviors over nondeterministic domains. Specifically, we tackle the problem of synthesizing strategies that guarantee the successful execution of a Golog program δ within a nondeterministic basic action theory (NDBAT). Our approach constructs symbolic program graphs that capture the control flow of δ independently of the domain, enabling reactive synthesis via their cross product with the domain model. We formally relate graph-based transitions to standard Golog semantics and show how our framework modularly separates agent and environment behaviors. This flexibility supports strategic reasoning and synthesis in complex nondeterministic settings, with programs acting as independent controllers for both agent and environment.

@article{de2025strategic,
title = {Strategic Reasoning over Golog Programs in the Nondeterministic Situation Calculus--Extended Abstract},
author = {De Giacomo, Giuseppe and Lesp{\'e}rance, Yves and Mancanelli, Matteo}, year = {2025}
}

[c3]

Situation Calculus Temporally Lifted Abstractions for Generalized Planning

Giuseppe De Giacomo, Yves Lespérance, Matteo Mancanelli

Proc. of the 39th Annual AAAI Conference on Artificial Intelligence (AAAI 2025), Philadelphia, PA, USA.

[abstract] [doi] [paper] [bibtex]

We present a new formal framework for generalized planning (GP) based on the situation calculus extended with LTL constraints. The GP problem is specified by a first-order basic action theory whose models are the problem instances. This low-level theory is then abstracted into a high-level propositional nondeterministic basic action theory with a single model. A refinement mapping relates the two theories. LTL formulas are used to specify the temporally extended goals as well as assumed trace constraints. If all LTL trace constraints hold at the low level and the high-level model can simulate all the low-level models with respect to the mapping, we say that we have a temporally lifted abstraction. We prove that if we have such an abstraction and the agent has a strategy to achieve a LTL goal under some trace constraints at the abstract level, then there exists a refinement of the strategy to achieve the refinement of the goal at the concrete level. We use LTL synthesis to generate the strategy at the abstract level. We illustrate our approach by synthesizing a program that solves a data structure manipulation problem.

@inproceedings{de2025situation,
title = {Situation Calculus Temporally Lifted Abstractions for Generalized Planning},
author = {De Giacomo, Giuseppe and Lesp{\'{e}}rance, Yves and Mancanelli, Matteo},
booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},
volume = {39},
number = {14},
pages = {14848--14857},
year = {2025}
}

[c2]

Situation Calculus Temporally Lifted Abstractions for Generalized Planning - Extended Abstract

Giuseppe De Giacomo, Yves Lespérance, Matteo Mancanelli

Workshop on Symbolic and Neuro-Symbolic Architectures for Intelligent Robotics Technology (SYNERGY 2024), co-located with KR 2024, Hanoi, Vietnam.

[abstract] [paper] [bibtex]

We present a new formal framework for generalized planning (GP) based on the situation calculus extended with LTL constraints. The GP problem is specified by a first-order basic action theory whose models are the problem instances. This low-level theory is then abstracted into a high-level propositional nondeterministic basic action theory with a single model. A refinement mapping relates the two theories. LTL formulas are used to specify the temporally extended goals as well as assumed trace constraints. If all LTL trace constraints hold at the low level and the high-level model can simulate all the low-level models with respect to the mapping, we say that we have a temporally lifted abstraction. We prove that if we have such an abstraction and the agent has a strategy to achieve a LTL goal under some trace constraints at the abstract level, then there exists a refinement of the strategy to achieve the refinement of the goal at the concrete level. We use LTL synthesis to generate the strategy at the abstract level. We illustrate our approach by synthesizing a program that solves a data structure manipulation problem.

@inproceedings{DBLP:conf/kodis/GiacomoLM24,
author = {Giuseppe De Giacomo and Yves Lesp{\'{e}}rance and Matteo Mancanelli},
editor = {Luc{\'{\i}}a G{\'{o}}mez {\'{A}}lvarez and Jonas Haldimann and Jesse Heyninck and Srdjan Vesic and Francesco Fabiano and Marcello Balduccini},
title = {Situation Calculus Temporally Lifted Abstractions for Generalized Planning - Extended Abstract},
booktitle = {Joint Proceedings of the Joint Workshop on Knowledge Diversity and Cognitive Aspects of {KR} and the Workshop on Symbolic and Neuro-Symbolic Architectures for Intelligent Robotics Technology (KoDis-CAKR-SYNERGY 2024) co-located with the 21st International Conference on Principles of Knowledge Representation and Reasoning {(KR} 2024), Hanoi, Vietnam, November 2-8, 2024},
series = {{CEUR} Workshop Proceedings},
volume = {3876},
publisher = {CEUR-WS.org},
year = {2024},
}

[c1]

PHYDI: Initializing Parameterized Hypercomplex Neural Networks as Identity Functions

Matteo Mancanelli, Eleonora Grassucci, Aurelio Uncini, Danilo Comminiello

IEEE International Workshop on Machine Learning for Signal Processing (MLSP 2023), Rome, Italy.

[abstract] [paper] [code] [bibtex] Top 5% Outstanding Paper

Neural models based on hypercomplex algebra systems are growing and prolificating for a plethora of applications, ranging from computer vision to natural language processing. Hand in hand with their adoption, parameterized hypercomplex neural networks (PHNNs) are growing in size and no techniques have been adopted so far to control their convergence at a large scale. In this paper, we study PHNNs convergence and propose parameterized hypercomplex identity initialization (PHYDI), a method to improve their convergence at different scales, leading to more robust performance when the number of layers scales up, while also reaching the same performance with fewer iterations. We show the effectiveness of this approach in different benchmarks and with common PHNNs with ResNets- and Transformer-based architecture.

@inproceedings{mancanelli2023MLSP,
author={Mancanelli, Matteo and Grassucci, Eleonora and Uncini, Aurelio and Comminiello, Danilo},
booktitle={2023 IEEE 33rd International Workshop on Machine Learning for Signal Processing (MLSP)},
title={{PHYDI: I}nitializing Parameterized Hypercomplex Neural Networks as Identity Functions},
year={2023},
organization={IEEE},
pages={1--6},
doi={10.1109/MLSP55844.2023.10285926}
}