REVIEW 64 references
Proof-Carrying Neuro-Symbolic Code
T0 review · reviewed 2026-08-16 · deepseek-v4-flash
Pith's one-line read The paper defines proof-carrying neuro-symbolic code, a research program for delivering neural-network-containing software with formal safety proofs, and reviews early tools and challenges.
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
The rest of the paper lists three challenges. First, existing proof assistants like Agda are good at high-level proofs but bad at verifying neural networks; specialized solvers like Marabou are the opposite. A compiler called Vehicle tries to connect the two. Second, if you trust a neural network solver, you need evidence you can check independently; the paper describes a proof checker for Marabou that was itself verified in the interactive prover Imandra. Third, specifications written in logic must be turned into loss functions for training, which requires a theory of differentiable logics. The paper reports early progress and open issues.
This is not a paper with a new theorem or experiment. It is an invited talk written as a position piece. Its value is in naming a research area and showing that existing tools already cover parts of it.
Extended reading notes
Core claim
In Section 2 the paper states: 'writing proof-carrying neuro-symbolic code amounts to writing a program s(u◦f◦e) and completing a proof as in equations (1) - (3).' If correct, neuro-symbolic programs can be delivered with end-to-end safety proofs by discharging three lemmas: a network property Ξ, a solution property Φ, and a program property Ψ.
Load-bearing premise
The framework assumes that for any neuro-symbolic program s(u◦f◦e) and any desired property Ψ, there exist a network property Ξ and a solution property Φ such that the three lemmas (1)-(3) hold, in particular the lifting lemma ∀g. Ξ(g) ⇒ Φ(u◦g◦e). Section 2 introduces this decomposition as the definition of proof-carrying neuro-symbolic code, but the paper does not prove the lemmas for the car example or identify conditions on e, u, and s under which the decomposition is complete. If the decomposition is not generally achievable, the framework only applies to specially structured programs.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Assumptions & free parameters
assumptions (3)
- domain assumption For any neuro-symbolic program s(u◦f◦e) and property Ψ, there exist Ξ and Φ such that lemmas (1)-(3) hold.
- domain assumption Neural network solvers such as Marabou are sound and can produce proof certificates; the Imandra-based checker is correct.
- domain assumption Robustness and similar universal properties cannot be learned from data alone and require formal specification.
invented entities (1)
-
proof-carrying neuro-symbolic code (term/framework)
Cite this review
Pith. "Pith review of Proof-Carrying Neuro-Symbolic Code." pith.science (2026). https://pith.science/paper/GY5PMSYZ
@misc{pith2026250412031,
author = {Pith},
title = {Pith review of: Proof-Carrying Neuro-Symbolic Code},
year = {2026},
howpublished = {\url{https://pith.science/paper/GY5PMSYZ}},
note = {Machine review of arXiv:2504.12031}
}
read the original abstract
This invited paper introduces the concept of "proof-carrying neuro-symbolic code" and explains its meaning and value, from both the "neural" and the "symbolic" perspectives. The talk outlines the first successes and challenges that this new area of research faces.
Figures
Reference graph
Works this paper leans on
-
[1]
Abate,A.,Giacobbe,M.,Roy,D.:Stochasticomega-regularverificationandcontrol with supermartingales. In: Gurfinkel, A., Ganesh, V. (eds.) Computer Aided Ver- ification - 36th International Conference, CAV 2024, Montreal, QC, Canada, July 24-27, 2024, Proceedings, Part III. Lecture Notes in Computer Science, vol. 14683, pp. 395–419. Springer (2024). https://do...
-
[2]
Affeldt, R., Bruni, A., Bertot, Y., Cohen, C., Kerjean, M., Mahboubi, A., Rouh- ling, D., Roux, P., Sakaguchi, K., Stone, Z., Strub, P.Y., Théry, L.: Analysis li- brary compatible with mathematical components. Available athttps://github. com/math-comp/analysis (2017), latest version: 1.8.0 (2024-12-19)
work page 2017
-
[3]
In: Bertot, Y., Kutsia, T., Norrish, M
Affeldt, R., Bruni, A., Komendantskaya, E., Ślusarz, N., Stark, K.: Taming Dif- ferentiable Logics with Coq Formalisation. In: Bertot, Y., Kutsia, T., Norrish, M. (eds.) 15th International Conference on Interactive Theorem Proving (ITP 2024). Leibniz International Proceedings in Informatics (LIPIcs), vol. 309, pp. 4:1–4:19. Schloss Dagstuhl – Leibniz-Zent...
-
[4]
In: 30th International Conference on Types for Proofs and Programs TYPES 2024–Abstracts
Affeldt, R., Bruni, A., Roux, P., Saikawa, T.: Yet another formal theory of probabil- ities (with an application to random sampling). In: 30th International Conference on Types for Proofs and Programs TYPES 2024–Abstracts. p. 11 (2024)
work page 2024
-
[5]
Affeldt, R., Cohen, C.: Measure construction by extension in dependent type theory with application to integration. J. Autom. Reason. 67(3), 28 (2023). https://doi.org/10.1007/S10817-023-09671-5, https://doi.org/10. 1007/s10817-023-09671-5
-
[6]
In: Bertot, Y., Kutsia, T., Norrish, M
Affeldt, R., Stone, Z.: A comprehensive overview of the lebesgue differentiation theorem in coq. In: Bertot, Y., Kutsia, T., Norrish, M. (eds.) 15th Interna- tional Conference on Interactive Theorem Proving, ITP 2024, September 9-14, 2024, Tbilisi, Georgia. LIPIcs, vol. 309, pp. 5:1–5:19. Schloss Dagstuhl - Leibniz- Zentrum für Informatik (2024). https://...
-
[7]
Atkey, R., Capucci, M., Komendantskaya, E., Mardare, R.: Quantitative predicate logic as a foundation for verified ml. ARIA grant (2024)
work page 2024
-
[8]
Annals of Pure and Ap- plied Logic 147(1), 23–47 (2007)
Baaz, M., Preining, N., Zach, R.: First-order gödel logics. Annals of Pure and Ap- plied Logic 147(1), 23–47 (2007). https://doi.org/https://doi.org/10.1016/ j.apal.2007.03.001, https://www.sciencedirect.com/science/article/pii/ S016800720700019X
work page 2007
Show all 64 references
-
[9]
Electronic Notes in Theoretical Informatics and Computer Sci- ence 3 (2023)
Bacci, G., Mardare, R., Panangaden, P., Plotkin, G.: Propositional logics for the lawvere quantale. Electronic Notes in Theoretical Informatics and Computer Sci- ence 3 (2023)
2023
-
[10]
In: Proc
Barbosa, H., Reynolds, A., Kremer, G., Lachnitt, H., Niemetz, A., Nötzli, A., Ozdemir, A., Preiner, M., Viswanathan, A., Viteri, S., Zohar, Y., Tinelli, C., Bar- rett, C.: Flexible Proof Production in an Industrial-Strength SMT Solver. In: Proc. 11th Int. Joint Conference on A...
2022
-
[11]
All about Proofs, Proofs for All55(1), 23–44 (2015)
Barrett, C., de Moura, L., Fontaine, P.: Proofs in Satisfiability Modulo Theories. All about Proofs, Proofs for All55(1), 23–44 (2015)
2015
-
[12]
Bourgain, J.: New classes of Lp-spaces, vol. 889. Springer (2006)
2006
-
[13]
arXiv preprint arXiv:2406.04936 (2024)
Capucci, M.: On quantifiers for quantitative reasoning. arXiv preprint arXiv:2406.04936 (2024)
2024
-
[14]
Carlini, N.: A complete list of all (arxiv) adversarial example papers (2019)
2019
-
[15]
In: Computer Aided Verification (CAV 2022)
Casadio, M., Komendantskaya, E., Daggitt, M.L., Kokke, W., Katz, G., Amir, G., Refaeli, I.: Neural network robustness as a verification property: A principled case study. In: Computer Aided Verification (CAV 2022). Lecture Notes in Computer Science, Springer (2022)
2022
-
[16]
In: International conference on computer aided verification
Casadio, M., Komendantskaya, E., Daggitt, M.L., Kokke, W., Katz, G., Amir, G., Refaeli, I.: Neural network robustness as a verification property: a principled case study. In: International conference on computer aided verification. pp. 219–231. Springer (2022) 10 E. Komendantskaya
2022
-
[17]
Foundations and Trends® in Programming Languages 7(3), 158–243 (2021)
Chaudhuri, S., Ellis, K., Polozov, O., Singh, R., Solar-Lezama, A., Yue, Y.: Neu- rosymbolic programming. Foundations and Trends® in Programming Languages 7(3), 158–243 (2021). https://doi.org/10.1561/2500000049, http://dx.doi. org/10.1561/2500000049
2021 doi
-
[18]
In: European Symposium on Programming Languages, ESOP 2025 (2025)
Cordeiro, L., Daggitt, M., Girard, J., Isac, O., Johnson, T., Katz, G., Komen- dantskaya, E., Manino, E., Sinkarovs, A., Wu, H.: Neural network verification is a programming language challenge. In: European Symposium on Programming Languages, ESOP 2025 (2025)
2025
-
[19]
In: Proceedings of the 12th ACM SIGPLAN International Conference on Certified Programs and Proofs
Daggitt, M.L., Atkey, R., Kokke, W., Komendantskaya, E., Arnaboldi, L.: Com- piling higher-order specifications to smt solvers: How to deal with rejection con- structively. In: Proceedings of the 12th ACM SIGPLAN International Conference on Certified Programs and Proofs. pp. 1...
2023
-
[20]
Daggitt, M.L., Kokke, W., Atkey, R.: Efficient compilation of expressive problem space specifications to neural network solvers (2024), https://arxiv.org/abs/ 2402.01353
2024
-
[21]
CoRRabs/2401.06379(2024)
Daggitt, M.L., Kokke, W., Atkey, R., Slusarz, N., Arnaboldi, L., Komendantskaya, E.: Vehicle: Bridging the embedding gap in the verification of neuro-symbolic pro- grams. CoRRabs/2401.06379(2024). https://doi.org/10.48550/ARXIV.2401. 06379, https://doi.org/10.48550/arXiv.2401.06379
-
[22]
arXiv preprint arXiv:2401.06379 (2024)
Daggitt, M.L., Kokke, W., Atkey, R., Slusarz, N., Arnaboldi, L., Komendantskaya, E.: Vehicle: Bridging the embedding gap in the verification of neuro-symbolic pro- grams. arXiv preprint arXiv:2401.06379 (2024)
2024 arXiv
-
[23]
In: Narodytska, N., Amir, G., Katz, G., Isac, O
Daggitt, M.L., Kokke, W., Komendantskaya, E., Atkey, R., Arnaboldi, L., Slusarz, N., Casadio, M., Coke, B., Lee, J.: The vehicle tutorial: Neural network verifica- tion with vehicle. In: Narodytska, N., Amir, G., Katz, G., Isac, O. (eds.) Proceed- ings of the 6th Workshop on F...
2023 doi
-
[24]
Dalrymple, D.: Safeguarded ai: constructing guaranteed safety (2024), programme Thesis
2024
- [25]
-
[26]
In: International Conference on Machine Learning
Fischer, M., Balunovic, M., Drachsler-Cohen, D., Gehr, T., Zhang, C., Vechev, M.: Dl2: training and querying neural networks with logic. In: International Conference on Machine Learning. pp. 1931–1941. PMLR (2019)
2019
-
[27]
In: Chaudhuri, K., Salakhutdinov, R
Fischer, M., Balunovic, M., Drachsler-Cohen, D., Gehr, T., Zhang, C., Vechev, M.T.: DL2: training and querying neural networks with logic. In: Chaudhuri, K., Salakhutdinov, R. (eds.) Proceedings of the 36th International Conference on Ma- chine Learning, ICML 2019, 9-15 June 2...
2019
-
[28]
Flinkow, T., Pearlmutter, B.A., Monahan, R.: Comparing differentiable logics for learning with logical constraints (2024),https://arxiv.org/abs/2407.03847
2024 arXiv
-
[29]
In: CADE
Fulton, N., et al.: KeYmaeraX: An axiomatic tactical theorem prover for hybrid systems. In: CADE. pp. 527–538 (2015). https://doi.org/10.1007/ 978-3-319-21401-6_36
2015
-
[30]
Elsevier (2007) Proof-Carrying Neuro-Symbolic Code 11
Galatos, N., Jipsen, P., Ono, T.K.H.: Residuated Lattices: An Algebraic Glimpse at Substructural Logics. Elsevier (2007) Proof-Carrying Neuro-Symbolic Code 11
2007
-
[31]
In: Proc
Gehr, T., Mirman, M., Drachsler-Cohen, D., Tsankov, P., Chaudhuri, S., Vechev, M.: AI2: Safety and Robustness Certification of Neural Networks with Abstract Interpretation. In: Proc. 39th IEEE Symposium on Security and Privacy (SP). pp. 3–18 (2018)
2018
-
[32]
arXiv preprint arXiv:2206.03044 (2022)
Girard-Satabin, J., Alberti, M., Bobot, F., Chihani, Z., Lemesle, A.: Caisar: A plat- form for characterizing artificial intelligence safety and robustness. arXiv preprint arXiv:2206.03044 (2022)
2022 arXiv
-
[33]
In: Thirty-First International Joint Conference on Artificial Intelli- gence (IJCAI-22)
Giunchiglia, E., Stoian, M.C., Lukasiewicz, T.: Deep learning with logical con- straints. In: Thirty-First International Joint Conference on Artificial Intelli- gence (IJCAI-22). pp. 5478–5485. International Joint Conferences on Artificial Intelligence Organization (7 2022).ht...
2022 doi
-
[34]
In: McMillan, K.L., Middeldorp, A., Voronkov, A
Heras, J., Komendantskaya, E., Johansson, M., Maclean, E.: Proof-pattern recogni- tion and lemma discovery in ACL2. In: McMillan, K.L., Middeldorp, A., Voronkov, A. (eds.) Logic for Programming, Artificial Intelligence, and Reasoning - 19th International Conference, LPAR-19, S...
2013
-
[35]
Ho, S., Fromherz, A., Protzenko, J.: Modularity, code specialization, and zero-cost abstractions for program verification. Proc. ACM Program. Lang.7(ICFP) (Aug 2023). https://doi.org/10.1145/3607844, https://doi.org/10.1145/3607844
2023 doi
-
[36]
In: Proc
Isac, O., Barrett, C., Zhang, M., Katz, G.: Neural Network Verification with Proof Production. In: Proc. 22nd Int. Conf. on Formal Methods in Computer-Aided Design (FMCAD). pp. 38–48 (2022)
2022
-
[37]
In: Proc
Jia, K., Rinard, M.: Exploiting Verified Neural Networks via Floating Point Nu- merical Error. In: Proc. 28th Int. Static Analysis Symposium (SAS). pp. 191–205 (2021)
2021
-
[38]
443–452 (07 2019)
Katz, G., Huang, D., Ibeling, D., Julian, K., Lazarus, C., Lim, R., Shah, P., Thakoor, S., Wu, H., Zeljić, A., Dill, D., Kochenderfer, M., Barrett, C.: The Marabou Framework for Verification and Analysis of Deep Neural Networks, pp. 443–452 (07 2019)
2019
-
[39]
Courier Corporation (2014)
Kazarinoff, N.D.: Analytic inequalities. Courier Corporation (2014)
2014
-
[40]
Kokke, W., Komendantskaya, E., Kienitz, D., Atkey, R., Aspinall, D.: Neu- ral networks, secure by construction - an exploration of refinement types. In: d. S. Oliveira, B.C. (ed.) Programming Languages and Systems - 18th Asian Symposium, APLAS 2020, Fukuoka, Japan, November 30...
2020 doi
-
[41]
NeurIPS 2018 tutorial (2018), available athttps://adversarial-ml-tutorial.org/
Kolter, Z., Madry, A.: Adversarial robustness—theory and practice. NeurIPS 2018 tutorial (2018), available athttps://adversarial-ml-tutorial.org/
2018
-
[42]
In: Kaliszyk, C., Lüth, C
Komendantskaya, E., Heras, J., Grov, G.: Machine learning in proof general: Interfacing interfaces. In: Kaliszyk, C., Lüth, C. (eds.) Proceedings 10th Inter- national Workshop On User Interfaces for Theorem Provers, UITP 2012, Bre- men, Germany, July 11th, 2012. EPTCS, vol. 11...
2012 doi
-
[43]
Lemesle, A., Lehmann, J., Gall, T.L.: Neural network verification with pyrat (2024), https://arxiv.org/abs/2410.23903
2024 arXiv
-
[44]
Communications of the ACM 52(7), 107–115 (2009) 12 E
Leroy, X.: Formal Verification of a Realistic Compiler. Communications of the ACM 52(7), 107–115 (2009) 12 E. Komendantskaya
2009
-
[45]
Mandal, U., Amir, G., Wu, H., Daukantas, I., Newell, F.L., Ravaioli, U.J., Meng, B., Durling, M., Ganai, M., Shim, T., Katz, G., Barrett, C.: Formally verifying deep reinforcement learning controllers with lyapunov barrier certificates (2024), https://arxiv.org/abs/2405.14058
2024 arXiv
-
[46]
Manginas, V., Manginas, N., Stevinson, E., Varghese, S., Katzouris, N., Paliouras, G., Lomuscio, A.: A scalable approach to probabilistic neuro-symbolic verification (2025), https://arxiv.org/abs/2502.03274
2025 arXiv
-
[47]
Marulanda-Giraldo, J.M., Komendantskaya, E., Bruni, A., Affeldt, R., Capucci, M.: Quantifiers for quantitative logics in rocq: a new project description (2025), a draft
2025
-
[48]
Metcalfe, G., Olivetti, N., Gabbay, D.M.: Proof theory for fuzzy logics, vol. 36. Springer Science & Business Media (2008)
2008
-
[49]
Communications of the ACM54(9), 69–77 (2011)
de Moura, L., Bjørner, N.: Satisfiability Modulo Theories: Introduction and Appli- cations. Communications of the ACM54(9), 69–77 (2011)
2011
-
[50]
In: Proc
Necula, G.: Proof-carrying code. In: Proc. 24th Symposium on Principles of Pro- gramming Languages (POPL). pp. 106–119 (1997)
1997
- [51]
-
[52]
In: Proc
Passmore, G., Cruanes, S., Ignatovich, D., Aitken, D., Bray, M., Kagan, E., Kani- shev, K., Maclean, E., Mometto, N.: The Imandra Automated Reasoning System (System Description). In: Proc. 10th Int. Joint Conf. Automated Reasoning (IJ- CAR). pp. 464–471 (2020)
2020
-
[53]
Springer, Cham (2018)
Platzer, A.: Logical Foundations of Cyber-Physical Systems. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-63588-0
2018 doi
-
[55]
In: Piskac, R., Voronkov, A
Slusarz, N., Komendantskaya, E., Daggitt, M.L., Stewart, R.J., Stark, K.: Logic of differentiable logics: Towards a uniform semantics of DL. In: Piskac, R., Voronkov, A. (eds.) LPAR 2023: Proceedings of 24th International Conference on Logic for Programming, Artificial Intelli...
2023 doi
-
[56]
Szegedy, C., Zaremba, W., Sutskever, I., Bruna, J., Erhan, D., Goodfellow, I., Fergus, R.: Intriguing properties of neural networks (2014)
2014
-
[57]
Team, M.C.: Mathematical components library.https://github.com/math-comp/ math-comp (2007)
2007
-
[58]
In: NeurIPS 2024 (2024), http://papers.nips.cc/paper_files/paper/2024/hash/ 031b5fd7d847f69ed33378a9a1117b4b-Abstract-Conference.html
Teuber, S., Mitsch, S., Platzer, A.: Provably safe neural net- work controllers via differential dynamic logic. In: NeurIPS 2024 (2024), http://papers.nips.cc/paper_files/paper/2024/hash/ 031b5fd7d847f69ed33378a9a1117b4b-Abstract-Conference.html
2024
-
[59]
Journal of the Operational Research Society (1996)
Vanderbei, R.: Linear Programming: Foundations and Extensions. Journal of the Operational Research Society (1996)
1996
-
[60]
Advances in Neural Information Processing Sys- tems 34, 29909–29921 (2021) Proof-Carrying Neuro-Symbolic Code 13
Wang, S., Zhang, H., Xu, K., Lin, X., Jana, S., Hsieh, C.J., Kolter, J.Z.: Beta- crown: Efficient bound propagation with per-neuron split constraints for neural network robustness verification. Advances in Neural Information Processing Sys- tems 34, 29909–29921 (2021) Proof-Ca...
2021
-
[61]
In: Proc
Wu, H., Isac, O., Zeljić, A., Tagomori, T., Daggitt, M., Kokke, W., Refaeli, I., Amir, G., Julian, K., Bassan, S., Huang, P., Lahav, O., Wu, M., Zhang, M., Komen- dantskaya, E., Katz, G., Barrett, C.: Marabou 2.0: A Versatile Formal Analyzer of Neural Networks. In: Proc. 36th ...
2024
-
[62]
In: Computer Aided Verification (CAV) (2024)
Wu, H., Isac, O., Zeljic, A., Tagomori, T., Daggitt, M.L., Kokke, W., Refaeli, I., Amir, G., Julian, K., Bassan, S., Huang, P., Lahav, O., Wu, M., Zhang, M., Komendantskaya, E., Katz, G., Barrett, C.W.: Marabou 2.0: A Versatile Formal Analyzer of Neural Networks. In: Computer ...
2024
- [63]
-
[64]
In: International Conference on Learning Representations (2021), https://api.semanticscholar.org/CorpusID:235613486
Zombori,D.,Bánhelyi,B.,Csendes,T.,Megyeri,I.,Jelasity,M.:Foolingacomplete neural network verifier. In: International Conference on Learning Representations (2021), https://api.semanticscholar.org/CorpusID:235613486
2021
-
[406]
Springer (2013).https://doi.org/10.1007/978-3-642-45221-5_27 , https: //doi.org/10.1007/978-3-642-45221-5_27
2013 doi
Reviewed August 16, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.