Reproduce “Deep Arguing” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Adam Gould, Francesca Toni

Deep Arguing introduces a neurosymbolic approach that combines deep neural networks with formal argumentation theory to deliver interpretable, case-based explanations for model predictions. The framework achieves competitive classification performance on various tabular and imaging datasets while providing transparent reasoning through learned argumentative structures.

Key insights

  • Deep learning can be effectively integrated with formal argumentation to learn explicit, case-based reasoning structures directly from data, providing faithful explanations.
  • Differentiable argumentation semantics allow for gradient-based optimization of neurosymbolic models, enabling joint learning of feature representations and argumentative interactions.
  • Structural constraints applied during training guide the model to generate more interpretable and stable argumentation graphs, enhancing both explanation quality and predictive performance.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.