pith. machine review for the scientific record. sign in

arxiv: 2604.24612 · v1 · submitted 2026-04-27 · 💻 cs.AI · cs.LO· math.CT· math.LO

Recognition: unknown

NeSyCat: A Monad-Based Categorical Semantics of the Neurosymbolic ULLER Framework

Daniel Romero Schellhorn , Till Mossakowski
Authors on Pith no claims yet

Pith reviewed 2026-05-08 03:32 UTC · model grok-4.3

classification 💻 cs.AI cs.LOmath.CTmath.LO
keywords ULLERneurosymbolic AIcategorical semanticsmonadsfirst-order logicLogic Tensor Networksunified semantics
0
0 comments X

The pith

The classical, fuzzy, and probabilistic semantics of ULLER arise as instances of one monad-based categorical framework.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

ULLER supplies one first-order logic syntax that different neurosymbolic systems can interpret classically, fuzzily, or probabilistically. The paper shows these three interpretations all fit inside a single categorical construction that uses monads, the structures that capture side effects such as uncertainty or nondeterminism in programming. This common structure makes it possible to add further interpretations or to translate formulas from one semantics to another while keeping the original behavior intact. A reader would care because it turns what look like separate reasoning engines into interchangeable modules inside one implementation.

Core claim

The paper shows that the semantic rules for classical, fuzzy, and probabilistic interpretations of ULLER's syntax can be captured uniformly as instances of a monad-based framework in category theory. This common framework supports the modular addition of new semantics, such as extending Logic Tensor Networks with generalized quantification over arbitrary domains via an extension of the Giry monad, and enables systematic translations between the different semantics.

What carries the argument

Monad-based categorical semantics, in which each of the three interpretations is obtained by equipping the same logic syntax with a different monad that encodes the corresponding notion of truth or probability.

If this is right

  • New semantic interpretations can be introduced by defining appropriate monads without changing the shared syntax.
  • Translations between classical, fuzzy, and probabilistic readings become definable as monad morphisms.
  • Modular implementations of ULLER become feasible in programming languages that support monads, such as Python and Haskell.
  • Generalized quantification can be added to systems like Logic Tensor Networks and applied to infinite domains.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same monadic construction could be reused to combine ULLER with other uncertainty calculi by supplying the corresponding monad.
  • Switching semantics inside a running neurosymbolic system might become a matter of composing different monads rather than rewriting the inference engine.
  • Formal properties proved inside one semantics could be lifted to others once the monad morphisms are established.

Load-bearing premise

The original classical, fuzzy, and probabilistic semantic rules of ULLER can be faithfully embedded into a monad-based framework without altering their core behavior or losing distinguishing properties.

What would settle it

A concrete ULLER knowledge base evaluated under the original semantic rules yields different results from the same base evaluated inside the monad-based framework for at least one formula.

Figures

Figures reproduced from arXiv: 2604.24612 by Daniel Romero Schellhorn, Till Mossakowski.

Figure 1
Figure 1. Figure 1: Argmax transformation flowchart for the traffic light example from [ view at source ↗
read the original abstract

ULLER (Unified Language for LEarning and Reasoning) offers a unified first-order logic (FOL) syntax, enabling its knowledge bases to be used directly across a wide range of neurosymbolic systems. The original specification endows this syntax with three pairwise independent semantics: classical, fuzzy, and probabilistic, each accompanied by dedicated semantic rules. We show that these seemingly disparate semantics are all instances of one categorical framework based on monads, the very construct that models side effects in functional programming. This enables the modular addition of new semantics and systematic translations between them. As example, we outline the addition of generalised quantification in Logic Tensor Networks (LTN) to arbitrary (also infinite) domains by extending the Giry monad to probability spaces. In particular, our approach allows a modular implementation of ULLER in Python and Haskell, of which we have published initial versions on GitHub.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces NeSyCat, a monad-based categorical semantics for the ULLER framework's unified first-order logic syntax. It claims that the original classical, fuzzy, and probabilistic semantics are all instances of a single monadic framework, enabling modular extensions of new semantics and systematic translations between them. As a concrete case, it outlines extending the Giry monad to support generalized quantification in Logic Tensor Networks over arbitrary (including infinite) domains, and notes initial Python and Haskell implementations published on GitHub.

Significance. If the monadic embeddings are shown to be faithful, the result would supply a unifying categorical lens for neurosymbolic reasoning systems, supporting modularity and cross-semantics translations in a principled way. The published code repositories constitute a concrete strength for reproducibility and practical verification.

major comments (2)
  1. [§4] §4 (main unification result): the central claim that the classical, fuzzy, and probabilistic ULLER rules are instances of the monad framework requires explicit preservation theorems showing that the monadic interpretations coincide with the pre-existing semantic rules on all formulas (preserving bivalence, t-norm values, and probability measures exactly). Without such derivations, the 'instances' statement remains an assertion rather than a verified embedding.
  2. [§5.2] §5.2 (LTN/Giry extension): the outline of the Giry-monad extension for infinite domains must address measurability and normalization conditions to ensure the probabilistic behavior is unaltered; any implicit restriction on the class of measures would constitute a silent change to the original ULLER probabilistic semantics and undermine the unification claim.
minor comments (2)
  1. [§3] The notation for monad operations (bind, return, etc.) should be aligned with standard references in the category-theory literature to improve readability for readers outside the immediate subfield.
  2. [Figure 2] Figure 2 (or equivalent diagram of the monad embeddings) would benefit from explicit labels indicating which arrows correspond to the classical, fuzzy, and probabilistic cases.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and insightful comments, which help clarify the presentation of our unification result. We address each major comment below and will incorporate the suggested clarifications and proofs in the revised manuscript.

read point-by-point responses

  1. Referee: [§4] §4 (main unification result): the central claim that the classical, fuzzy, and probabilistic ULLER rules are instances of the monad framework requires explicit preservation theorems showing that the monadic interpretations coincide with the pre-existing semantic rules on all formulas (preserving bivalence, t-norm values, and probability measures exactly). Without such derivations, the 'instances' statement remains an assertion rather than a verified embedding.

    Authors: We agree that explicit preservation theorems are required for a fully rigorous verification. The monadic semantics in §4 are defined via the unit and bind operations so that they replicate the original rules by construction for atomic formulas and are extended homomorphically; however, to make this fully explicit, the revised version will add dedicated preservation theorems in §4. These will prove by induction on formula structure that the monadic interpretation exactly recovers bivalence in the classical case, the chosen t-norm values in the fuzzy case, and the probability measures in the probabilistic case, for every formula. revision: yes

  2. Referee: [§5.2] §5.2 (LTN/Giry extension): the outline of the Giry-monad extension for infinite domains must address measurability and normalization conditions to ensure the probabilistic behavior is unaltered; any implicit restriction on the class of measures would constitute a silent change to the original ULLER probabilistic semantics and undermine the unification claim.

    Authors: We appreciate this observation. The Giry-monad extension in §5.2 will be expanded to explicitly state the measurability requirements (i.e., that all predicates and quantifiers are interpreted via measurable functions on the underlying measurable space) and to confirm that normalization is preserved under the monadic bind operation. The revised text will specify that we work with the standard Giry monad on Polish spaces (or more generally on measurable spaces where the integrals remain well-defined) and will verify that no additional restrictions are imposed beyond those already present in the original ULLER probabilistic semantics, thereby preserving the unification. revision: yes

Circularity Check

0 steps flagged

Monad-based unification of ULLER semantics is self-contained with no circular reduction

full rationale

The paper constructs a monad-based categorical framework and demonstrates that the pre-existing classical, fuzzy, and probabilistic semantics of ULLER are instances within it, preserving their original rules. This is presented as a modular embedding and unification rather than a derivation that reduces to fitted parameters, self-definitions, or load-bearing self-citations. The Giry monad extension for infinite domains in LTN is an illustrative addition of a new semantics, not a forced prediction from the core claim. No equation or step equates a result to its own input by construction; the framework is shown to recover the original behaviors without alteration, making the derivation independent and externally verifiable against the stated ULLER rules.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper relies on standard category theory (monads as models of side effects) and the original ULLER semantics as background; no free parameters or new invented entities are mentioned in the abstract. The monad framework itself is the main contribution rather than an ad-hoc postulate.

axioms (1)
  • standard math Monads model side effects in functional programming and can be used to structure semantics
    Invoked in the abstract as the basis for unifying the three semantics.

pith-pipeline@v0.9.0 · 5459 in / 1224 out tokens · 49936 ms · 2026-05-08T03:32:41.053160+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

43 extracted references · 26 canonical work pages

  1. [1]

    Ad\' a mek, H

    J. Ad\' a mek, H. Herrlich, and G. Strecker. Abstract and Concrete Categories. Wiley, New York, 1990

    1990

  2. [2]

    S. Awodey. Category Theory. Number 52 in Oxford Logic Guides. Oxford Univ. Press, Oxford, 2. ed edition, 2010. ISBN 978-0-19-958736-0 978-0-19-923718-0

    2010

  3. [3]

    Badreddine and M

    S. Badreddine and M. Spranger. Extending real logic with aggregate functions. In A. S. d'Avila Garcez and E. Jim \' e nez - Ruiz, editors, Proceedings of the 15th International Workshop on Neural-Symbolic Learning and Reasoning as part of the 1st International Joint Conference on Learning & Reasoning (IJCLR 2021), Virtual conference, October 25-27, 2021 ,...

    2021

  4. [4]

    Badreddine, A

    S. Badreddine, A. S. d'Avila Garcez, L. Serafini, and M. Spranger. Logic tensor networks. Artif. Intell., 303: 0 103649, 2022. doi:10.1016/J.ARTINT.2021.103649. URL https://doi.org/10.1016/j.artint.2021.103649

    work page doi:10.1016/j.artint.2021.103649 2022

  5. [5]

    Disintegration and Bayesian inversion via string diagrams

    K. Cho and B. Jacobs. Disintegration and Bayesian inversion via string diagrams. Mathematical Structures in Computer Science, 29 0 (7): 0 938--971, 2019. doi:10.1017/S0960129518000488

    work page doi:10.1017/s0960129518000488 2019

  6. [6]

    Coecke, M

    B. Coecke, M. Sadrzadeh, and S. Clark. ( DisCoCat ) Mathematical Foundations for a Compositional Distributional Model of Meaning , Mar. 2010

    2010

  7. [7]

    B. Fong, D. I. Spivak, and R. Tuy \'e ras. Backprop as functor: A compositional perspective on supervised learning. In 2019 34th Annual ACM/IEEE Symposium on Logic in Computer Science (LICS), pages 1--13. IEEE, 2019. doi:10.1109/LICS.2019.8785665

    work page doi:10.1109/lics.2019.8785665 2019

  8. [8]

    T. Fritz. A synthetic approach to Markov kernels, conditional independence and theorems on sufficient statistics. Advances in Mathematics, 370: 0 107239, 2020. doi:10.1016/j.aim.2020.107239

    work page doi:10.1016/j.aim.2020.107239 2020

  9. [9]

    Fritz and P

    T. Fritz and P. Perrone. Bimonoidal structure of probability monads. Electronic Notes in Theoretical Computer Science, 341: 0 121--149, 2018. doi:10.1016/j.entcs.2018.11.007

    work page doi:10.1016/j.entcs.2018.11.007 2018

  10. [10]

    Gavranovi \'c , P

    B. Gavranovi \'c , P. Lessard, A. Dudzik, T. von Glehn, J. G. M. Ara \'u jo, and P. Veli c kovi \'c . Position: Categorical Deep Learning is an Algebraic Theory of All Architectures , June 2024

    2024

  11. [11]

    J.-Y. Girard. Linear Logic : Its syntax and semantics. In J.-Y. Girard, Y. Lafont, and L. Regnier, editors, Advances in Linear Logic , pages 1--42. Cambridge University Press, 1 edition, June 1995. ISBN 978-0-521-55961-4 978-0-511-62915-0. doi:10.1017/CBO9780511629150.002

    work page doi:10.1017/cbo9780511629150.002 1995

  12. [12]

    M. Giry. A categorical approach to probability theory. In Categorical Aspects of Topology and Analysis, volume 915 of Lecture Notes in Mathematics, pages 68--85, Berlin, Heidelberg, 1982. Springer

    1982

  13. [14]

    Harel, D

    D. Harel, D. Kozen, and J. Tiuryn. Dynamic logic. ACM SIGACT News, 32 0 (1): 0 66--69, 2001

    2001

  14. [15]

    Heunen, O

    C. Heunen, O. Kammar, S. Staton, and H. Yang. A Convenient Category for Higher-Order Probability Theory . In 2017 32nd Annual ACM / IEEE Symposium on Logic in Computer Science ( LICS ) , pages 1--12, June 2017. doi:10.1109/LICS.2017.8005137

    work page doi:10.1109/lics.2017.8005137 2017

  15. [16]

    P. Hájek. Metamathematics of Fuzzy Logic . Trends in Logic . Springer Netherlands, Dordrecht, 1998. ISBN 978-1-4020-0370-7 978-94-011-5300-3. doi:10.1007/978-94-011-5300-3. URL http://link.springer.com/10.1007/978-94-011-5300-3

    work page doi:10.1007/978-94-011-5300-3 1998

  16. [17]

    B. Jacobs. From probability monads to commutative effectuses. Journal of Logical and Algebraic Methods in Programming, 94: 0 200--237, 2018. doi:10.1016/j.jlamp.2016.11.006

    work page doi:10.1016/j.jlamp.2016.11.006 2018

  17. [18]

    P. T. Johnstone. Sketches of an Elephant A Topos Theory Compendium . Oxford University PressOxford, Sept. 2002. ISBN 978-0-19-851598-2 978-1-383-02287-2. doi:10.1093/oso/9780198515982.001.0001

    work page doi:10.1093/oso/9780198515982.001.0001 2002

  18. [19]

    Jones and G

    C. Jones and G. Plotkin. A probabilistic powerdomain of evaluations. In Proceedings of the Fourth Annual Symposium on Logic in Computer Science (LICS), pages 186--195, 1989. doi:10.1109/LICS.1989.39173

    work page doi:10.1109/lics.1989.39173 1989

  19. [20]

    Springer New York, NY, 1978.doi: 10.1007/978-1-4757-4721-8

    S. Mac Lane. Categories for the Working Mathematician , volume 5 of Graduate Texts in Mathematics . Springer New York, New York, NY, 1978. ISBN 978-1-4419-3123-8 978-1-4757-4721-8. doi:10.1007/978-1-4757-4721-8

    work page doi:10.1007/978-1-4757-4721-8 1978

  20. [21]

    Reyes.First Order Categorical Logic: Model-Theoretical Methods in the Theory of Topoi and Related Categories, volume 611 ofLecture Notes in Mathematics

    M. Makkai and G. E. Reyes. First Order Categorical Logic: Model-Theoretical Methods in the Theory of Topoi and Related Categories, volume 611 of Lecture Notes in Mathematics. Springer, Berlin, Heidelberg, 1977. doi:10.1007/BFb0066201

    work page doi:10.1007/bfb0066201 1977

  21. [22]

    Manhaeve, S

    R. Manhaeve, S. Dumancic, A. Kimmig, T. Demeester, and L. D. Raedt. Neural probabilistic logic programming in DeepProbLog . Artif. Intell., 298: 0 103504, 2021. doi:10.1016/J.ARTINT.2021.103504. URL https://doi.org/10.1016/j.artint.2021.103504

    work page doi:10.1016/j.artint.2021.103504 2021

  22. [23]

    E. Moggi. Notions of computation and monads. Information and Computation, 93 0 (1): 0 55--92, July 1991. ISSN 08905401. doi:10.1016/0890-5401(91)90052-4

    work page doi:10.1016/0890-5401(91)90052-4 1991

  23. [24]

    Morettin, A

    P. Morettin, A. Passerini, and R. Sebastiani. Efficient Weighted Model Integration via SMT-Based Predicate Abstraction . In Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence , pages 720--728, Melbourne, Australia, Aug. 2017. International Joint Conferences on Artificial Intelligence Organization. doi:10.24963/ijcai.2017/100

    work page doi:10.24963/ijcai.2017/100 2017

  24. [25]

    Mossakowski, L

    T. Mossakowski, L. Schr \" o der, and S. Goncharov. A generic complete dynamic logic for reasoning about purity and effects. Formal Aspects Comput., 22 0 (3-4): 0 363--384, 2010. doi:10.1007/S00165-010-0153-4. URL https://doi.org/10.1007/s00165-010-0153-4

    work page doi:10.1007/s00165-010-0153-4 2010

  25. [26]

    G. Priest. An introduction to non-classical logic: From if to is. Cambridge University Press, 2008

    2008

  26. [27]

    Ramsey and A

    N. Ramsey and A. Pfeffer. Stochastic lambda calculus and monads of probability distributions. In Proceedings of the 29th ACM SIGPLAN-SIGACT Symposium on Principles of Programming Languages (POPL), pages 154--165, 2002. doi:10.1145/503272.503288

    work page doi:10.1145/503272.503288 2002

  27. [28]

    D. R. Schellhorn and T. Mossakowski. muller: A modular monad-based semantics of the neurosymbolic uller framework. In L. H. Gilpin, E. Giunchiglia, P. Hitzler, and E. van Krieken, editors, Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning, volume 284 of Proceedings of Machine Learning Research, pages 494--518. PMLR, ...

    2025

  28. [29]

    \'S cibior, Z

    A. \'S cibior, Z. Ghahramani, and A. D. Gordon. Practical probabilistic programming with monads. In Proceedings of the 2015 ACM SIGPLAN Symposium on Haskell, pages 165--176, 2015

    2015

  29. [30]

    Moss, Chris Heunen, and Zoubin Ghahramani

    A. \'S cibior , O. Kammar, M. Vakar, S. Staton, H. Yang, Y. Cai, K. Ostermann, S. K. Moss, C. Heunen, and Z. Ghahramani. Denotational validation of higher-order Bayesian inference. volume 2, 2018. doi:10.1145/3158148

    work page doi:10.1145/3158148 2018

  30. [31]

    \'S lusarz, E

    N. \'S lusarz, E. Komendantskaya, M. L. Daggitt, R. Stewart, and K. Stark. Logic of Differentiable Logics : Towards a Uniform Semantics of DL , Oct. 2023

    2023

  31. [32]

    L. D. Smet, P. Z. D. Martires, R. Manhaeve, G. Marra, A. Kimmig, and L. D. Raedt. Neural Probabilistic Logic Programming in Discrete-Continuous Domains , Mar. 2023

    2023

  32. [33]

    D. I. Spivak. Category Theory for the Sciences. The MIT Press, Cambridge, Massachusetts, 2014. ISBN 978-0-262-02813-4

    2014

  33. [34]

    S. Staton. Commutative semantics for probabilistic programming. In Programming Languages and Systems (ESOP 2017), volume 10201 of LNCS, pages 855--879. Springer, 2017. doi:10.1007/978-3-662-54434-1_32

    work page doi:10.1007/978-3-662-54434-1_32 2017

  34. [35]

    S. Staton. Probabilistic programs as measures. In G. Barthe, J.-P. Katoen, and A. Silva, editors, Foundations of Probabilistic Programming, pages 43--74. Cambridge University Press, 2020. doi:10.1017/9781108770750.003

    work page doi:10.1017/9781108770750.003 2020

  35. [36]

    Staton, H

    S. Staton, H. Yang, F. Wood, C. Heunen, and O. Kammar. Semantics for probabilistic programming: higher-order functions, continuous distributions, and soft constraints. In Proceedings of the 31st Annual ACM/IEEE Symposium on Logic in Computer Science (LICS), pages 525--534, 2016. doi:10.1145/2933575.2935313

    work page doi:10.1145/2933575.2935313 2016

  36. [37]

    J. Tobin. Embedded Domain-Specific Languages for Bayesian Modelling and Inference. PhD thesis, University of Auckland, 2018

    2018

  37. [38]

    Toumi and G

    A. Toumi and G. de Felice. Higher- Order DisCoCat ( Peirce-Lambek-Montague semantics). Electronic Proceedings in Theoretical Computer Science, 429: 0 130--145, Sept. 2025. ISSN 2075-2180. doi:10.4204/EPTCS.429.7

    work page doi:10.4204/eptcs.429.7 2025

  38. [39]

    Vakar, O

    M. Vakar, O. Kammar, and S. Staton. A domain theory for statistical probabilistic programming. Proceedings of the ACM on Programming Languages, 3 0 (POPL), 2019. doi:10.1145/3290349

    work page doi:10.1145/3290349 2019

  39. [40]

    Van Krieken, S

    E. Van Krieken, S. Badreddine, R. Manhaeve, and E. Giunchiglia. ULLER : A Unified Language for Learning and Reasoning . In T. R. Besold, A. d'Avila Garcez , E. Jimenez-Ruiz , R. Confalonieri, P. Madhyastha, and B. Wagner, editors, Neural- Symbolic Learning and Reasoning , volume 14979, pages 219--239. Springer Nature Switzerland, Cham, 2024. ISBN 978-3-03...

    work page doi:10.1007/978-3-031-71167-1_12 2024

  40. [41]

    van Krieken, P

    E. van Krieken, P. Minervini, E. M. Ponti, and A. Vergari. On the Independence Assumption in Neurosymbolic Learning , June 2024

    2024

  41. [42]

    van Krieken, P

    E. van Krieken, P. Minervini, E. Ponti, and A. Vergari. Neurosymbolic Reasoning Shortcuts under the Independence Assumption , July 2025

    2025

  42. [43]

    Varnai and D

    P. Varnai and D. V. Dimarogonas. On Robustness Metrics for Learning STL Tasks . In 2020 American Control Conference ( ACC ) , pages 5394--5399, July 2020. doi:10.23919/ACC45564.2020.9147692

    work page doi:10.23919/acc45564.2020.9147692 2020

  43. [44]

    Walicki and S

    M. Walicki and S. Meldal. Multialgebras, power algebras and complete calculi of identities and inclusions. In Workshop on the Specification of Abstract Data Types, pages 453--468. Springer, 1994

    1994