Reproduce “AutoGraphForge: Towards Automated Graph Theory Discovery” 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
Ján Pastorek
AutoGraphForge, a computational pipeline developed at Comenius University, integrates automated conjecturing, aggressive refutation, novelty filtering, formalization, and theorem proving for graph theory. The system generated 6,522 unique conjectures, with 48 deemed "promising" and two non-trivial, provable relations between the annihilation number and edge-cover number identified for specific graph classes.
Key insights
- Integrating all four stages of mathematical discovery (conjecturing, refuting, formalizing, proving) into a single, closed-loop computational pipeline provides a robust approach to mitigating common failure modes of prior systems.
- An adversarial, counterexample-guided generation strategy, where counterexamples continuously update the graph dataset, is crucial for producing generally applicable and refutation-hardened conjectures.
- The combination of extensive static refutation datasets, diverse active counterexample search algorithms, and a comprehensive novelty filter is essential for rigorously validating and ensuring the originality of machine-generated mathematical statements.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.