Learning and Reasoning with Model-Grounded Symbolic Artificial Intelligence Systems

Aniruddha Chattopadhyay, Raj Dandekar, Kaushik Roy
Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning, PMLR 284:957-976, 2025.

Abstract

Neurosymbolic artificial intelligence (AI) systems combine neural network and classical symbolic AI mechanisms to exploit the complementary strengths of large-scale, generalizable learning and robust, verifiable reasoning. Numerous classifications of neurosymbolic AI illustrate how these two components can be integrated in distinctly different ways. In this work, we propose reinterpreting instruction-tuned large language models as model-grounded symbolic AI systems—where natural language serves as the symbolic layer, and grounding is achieved through the model’s internal representation space. Within this framework, we investigate and develop novel learning and reasoning approaches that preserve structural similarities to traditional learning and reasoning paradigms. Comprehensive evaluations across complex mathematical reasoning procedures of varying difficulties provide insights into the effectiveness of our approach towards learning efficiency and reasoning reliability.

Cite this Paper

BibTeX
@InProceedings{pmlr-v284-chattopadhyay25a, title = {Learning and Reasoning with Model-Grounded Symbolic Artificial Intelligence Systems}, author = {Chattopadhyay, Aniruddha and Dandekar, Raj and Roy, Kaushik}, booktitle = {Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning}, pages = {957--976}, 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/chattopadhyay25a/chattopadhyay25a.pdf}, url = {https://proceedings.mlr.press/v284/chattopadhyay25a.html}, abstract = {Neurosymbolic artificial intelligence (AI) systems combine neural network and classical symbolic AI mechanisms to exploit the complementary strengths of large-scale, generalizable learning and robust, verifiable reasoning. Numerous classifications of neurosymbolic AI illustrate how these two components can be integrated in distinctly different ways. In this work, we propose reinterpreting instruction-tuned large language models as model-grounded symbolic AI systems—where natural language serves as the symbolic layer, and grounding is achieved through the model’s internal representation space. Within this framework, we investigate and develop novel learning and reasoning approaches that preserve structural similarities to traditional learning and reasoning paradigms. Comprehensive evaluations across complex mathematical reasoning procedures of varying difficulties provide insights into the effectiveness of our approach towards learning efficiency and reasoning reliability.} }
Endnote
%0 Conference Paper %T Learning and Reasoning with Model-Grounded Symbolic Artificial Intelligence Systems %A Aniruddha Chattopadhyay %A Raj Dandekar %A Kaushik Roy %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-chattopadhyay25a %I PMLR %P 957--976 %U https://proceedings.mlr.press/v284/chattopadhyay25a.html %V 284 %X Neurosymbolic artificial intelligence (AI) systems combine neural network and classical symbolic AI mechanisms to exploit the complementary strengths of large-scale, generalizable learning and robust, verifiable reasoning. Numerous classifications of neurosymbolic AI illustrate how these two components can be integrated in distinctly different ways. In this work, we propose reinterpreting instruction-tuned large language models as model-grounded symbolic AI systems—where natural language serves as the symbolic layer, and grounding is achieved through the model’s internal representation space. Within this framework, we investigate and develop novel learning and reasoning approaches that preserve structural similarities to traditional learning and reasoning paradigms. Comprehensive evaluations across complex mathematical reasoning procedures of varying difficulties provide insights into the effectiveness of our approach towards learning efficiency and reasoning reliability.
APA
Chattopadhyay, A., Dandekar, R. & Roy, K.. (2025). Learning and Reasoning with Model-Grounded Symbolic Artificial Intelligence Systems. Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning, in Proceedings of Machine Learning Research 284:957-976 Available from https://proceedings.mlr.press/v284/chattopadhyay25a.html.

Related Material