mlsanityData to 14 Oct 2026

Generating Symbolic Reasoning Problems with Transformer GANs

arXiv 2110.10054 ↗First posted 19 Oct 2021cs.LGcited by 5

Not on the map. Why?

Abstract (arXiv)

We study the capabilities of GANs and Wasserstein GANs equipped with Transformer encoders to generate sensible and challenging training data for symbolic reasoning domains. We conduct experiments on two problem domains where Transformers have been successfully applied recently: symbolic mathematics and temporal specifications in verification. Even without autoregression, our GAN models produce syntactically correct instances. We show that the generated data can be used as a substitute for real training data when training a classifier, and, especially, that training data can be generated from a dataset that is too small to be trained on directly. Using a GAN setting also allows us to alter the target distribution: We show that by adding a classifier uncertainty part to the generator objective, we obtain a dataset that is even harder to solve for a temporal logic classifier than our original dataset.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Builds on 45 of its 94 references that are in the map. First posted on arXiv 19 Oct 2021.

Builds on

45 of 94 references.

References · 45

  1. 2017Attention Is All You Needcited by 60,361
  2. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  3. 2017Neural Message Passing for Quantum Chemistrycited by 3,487
  4. 2019A Comprehensive Survey on Graph Neural Networkscited by 2,919
  5. 2017Improved Training of Wasserstein GANscited by 2,235
  6. 2017Wasserstein GANcited by 639
  7. 2020Generative Language Modeling for Automated Theorem Provingcited by 388
  8. 2016SeqGAN: Sequence Generative Adversarial Nets with Policy Gradientcited by 376
  9. 2019Analysing Mathematical Reasoning Abilities of Neural Modelscited by 335
  10. 2019Deep Learning for Symbolic Mathematicscited by 270
  11. 2018Learning a SAT Solver from Single-Bit Supervisioncited by 200
  12. 2021Generative Adversarial Transformerscited by 120
33 more
  1. 2019Learning to Prove Theorems via Interacting with Proof Assistantscited by 109
  2. 2018GAN Augmentation: Augmenting Training Data using Generative Adversarial Networkscited by 107
  3. 2017Long Text Generation via Adversarial Training with Leaked Informationcited by 100
  4. 2018MaskGAN: Better Text Generation via Filling in the______cited by 89
  5. 2016DeepMath - Deep Sequence Models for Premise Selectioncited by 81
  6. 2019Guiding High-Performance SAT Solvers with Unsat-Core Predictionscited by 62
  7. 2017Adversarial Feature Matching for Text Generationcited by 58
  8. 2019Graph Representations for Higher-Order Logic and Theorem Provingcited by 55
  9. 2020Teaching Temporal Logics to Neural Networkscited by 51
  10. 2017Adversarial Ranking for Language Generationcited by 50
  11. 2016GANS for Sequences of Discrete Elements with the Gumbel-softmax Distributioncited by 48
  12. 2018GamePad: A Learning Environment for Theorem Provingcited by 48
  13. 2017Deep Network Guided Proof Searchcited by 47
  14. 2018TacticToe: Learning to Prove with Tacticscited by 44
  15. 2020INT: An Inequality Benchmark for Evaluating Generalization in Theorem Provingcited by 44
  16. 2017Maximum-Likelihood Augmented Discrete Generative Adversarial Networkscited by 39
  17. 2021TacticZero: Learning to Prove Theorems from Scratch with Deep Reinforcement Learningcited by 37
  18. 2019Enhancing the Transformer with Explicit Relational Encoding for Math Problem Solvingcited by 35
  19. 2021Goal-Aware Neural SAT Solvercited by 28
  20. 2018Adversarial Text Generation via Feature-Mover's Distancecited by 27
  21. 2017HolStep: A Machine Learning Dataset for Higher-order Logic Theorem Provingcited by 26
  22. 2020First Neural Conjecturing Datasets and Experimentscited by 26
  23. 2016Learning Continuous Semantic Representations of Symbolic Expressionscited by 25
  24. 2018Can Neural Networks Understand Logical Entailment?cited by 24
  25. 2014Temporal Logics for Hyperpropertiescited by 19
  26. 2019Mathematical Reasoning in Latent Spacecited by 17
  27. 2021Neural Circuit Synthesis from Specification Patternscited by 17
  28. 2018Learning Heuristics for Quantified Boolean Formulas through Deep Reinforcement Learningcited by 15
  29. 2014Learning to Discover Efficient Mathematical Identitiescited by 9
  30. 2018BoSy: An Experimentation Framework for Bounded Synthesiscited by 8
  31. 2020IsarStep: a Benchmark for High-level Mathematical Reasoningcited by 7
  32. 2020Style Example-Guided Text Generation using Generative Adversarial Transformerscited by 6
  33. 2015The First Reactive Synthesis Competition (SYNTCOMP 2014)cited by 1

Built on by

Cited by year

Citing papers per year: 2022: 1; 2023: 2; 2024: 1; 2025: 1.

5 papers.

  1. 2022Autoformalization with Large Language Modelscited by 263
  2. 2023nl2spec: Interactively Translating Unstructured Natural Language to Temporal Logics with Large Language Modelscited by 60
  3. 2023Iterative Circuit Repair Against Formal Specificationscited by 9
  4. 2024NeuroSynt: A Neuro-symbolic Portfolio Solver for Reactive Synthesiscited by 3
  5. 2025Breaking the Myth: Can Small Models Infer Postconditions Too?cited by 1