T-ILR: a Neurosymbolic Integration for LTLf

Riccardo Andreoni, Andrei Buliga, Alessandro Daniele, Chiara Ghidini, Marco Montali, Massimiliano Ronzani
Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning, PMLR 284:252-265, 2025.

Abstract

State-of-the-art approaches for integrating symbolic knowledge with deep learning architectures have demonstrated promising results in static domains. However, methods to handle temporal logic specifications remain underexplored. The only existing approach relies on an explicit representation of a finite-state automaton corresponding to the temporal specification. Instead, we aim at directly injecting the temporal knowledge into the neural model without having to rely on a separate symbolic structure. Specifically, we propose a neurosymbolic framework designed to incorporate temporal logic specifications, expressed in Linear Temporal Logic over finite traces (LTLf), directly into deep learning architectures for sequence-based tasks. We extend the Iterative Local Refinement (ILR) neurosymbolic algorithm, leveraging the recent introduction of fuzzy LTLf interpretations. We name this proposed method Temporal Iterative Local Refinement (T-ILR). We assess T-ILR on an existing benchmark for temporal neurosymbolic architectures, consisting of the classification of image sequences in the presence of temporal knowledge. The results demonstrate improved accuracy and computational efficiency compared to the state-of-the-art method.

Cite this Paper

BibTeX
@InProceedings{pmlr-v284-andreoni25a, title = {T-ILR: a Neurosymbolic Integration for LTLf}, author = {Andreoni, Riccardo and Buliga, Andrei and Daniele, Alessandro and Ghidini, Chiara and Montali, Marco and Ronzani, Massimiliano}, booktitle = {Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning}, pages = {252--265}, year = {2025}, editor = {H. Gilpin, Leilani and Giunchiglia, Eleonora and Hitzler, Pascal and van Krieken, Emile}, volume = {284}, series = {Proceedings of Machine Learning Research}, month = {08--10 Sep}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v284/main/assets/andreoni25a/andreoni25a.pdf}, url = {https://proceedings.mlr.press/v284/andreoni25a.html}, abstract = {State-of-the-art approaches for integrating symbolic knowledge with deep learning architectures have demonstrated promising results in static domains. However, methods to handle temporal logic specifications remain underexplored. The only existing approach relies on an explicit representation of a finite-state automaton corresponding to the temporal specification. Instead, we aim at directly injecting the temporal knowledge into the neural model without having to rely on a separate symbolic structure. Specifically, we propose a neurosymbolic framework designed to incorporate temporal logic specifications, expressed in Linear Temporal Logic over finite traces (LTLf), directly into deep learning architectures for sequence-based tasks. We extend the Iterative Local Refinement (ILR) neurosymbolic algorithm, leveraging the recent introduction of fuzzy LTLf interpretations. We name this proposed method Temporal Iterative Local Refinement (T-ILR). We assess T-ILR on an existing benchmark for temporal neurosymbolic architectures, consisting of the classification of image sequences in the presence of temporal knowledge. The results demonstrate improved accuracy and computational efficiency compared to the state-of-the-art method.} }
Endnote
%0 Conference Paper %T T-ILR: a Neurosymbolic Integration for LTLf %A Riccardo Andreoni %A Andrei Buliga %A Alessandro Daniele %A Chiara Ghidini %A Marco Montali %A Massimiliano Ronzani %B Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning %C Proceedings of Machine Learning Research %D 2025 %E Leilani H. Gilpin %E Eleonora Giunchiglia %E Pascal Hitzler %E Emile van Krieken %F pmlr-v284-andreoni25a %I PMLR %P 252--265 %U https://proceedings.mlr.press/v284/andreoni25a.html %V 284 %X State-of-the-art approaches for integrating symbolic knowledge with deep learning architectures have demonstrated promising results in static domains. However, methods to handle temporal logic specifications remain underexplored. The only existing approach relies on an explicit representation of a finite-state automaton corresponding to the temporal specification. Instead, we aim at directly injecting the temporal knowledge into the neural model without having to rely on a separate symbolic structure. Specifically, we propose a neurosymbolic framework designed to incorporate temporal logic specifications, expressed in Linear Temporal Logic over finite traces (LTLf), directly into deep learning architectures for sequence-based tasks. We extend the Iterative Local Refinement (ILR) neurosymbolic algorithm, leveraging the recent introduction of fuzzy LTLf interpretations. We name this proposed method Temporal Iterative Local Refinement (T-ILR). We assess T-ILR on an existing benchmark for temporal neurosymbolic architectures, consisting of the classification of image sequences in the presence of temporal knowledge. The results demonstrate improved accuracy and computational efficiency compared to the state-of-the-art method.
APA
Andreoni, R., Buliga, A., Daniele, A., Ghidini, C., Montali, M. & Ronzani, M.. (2025). T-ILR: a Neurosymbolic Integration for LTLf. Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning, in Proceedings of Machine Learning Research 284:252-265 Available from https://proceedings.mlr.press/v284/andreoni25a.html.

Related Material