![[2602.16954] Neural Proposals, Symbolic Guarantees: Neuro-Symbolic Graph Generation with Hard Constraints](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.16954] Neural Proposals, Symbolic Guarantees: Neuro-Symbolic Graph Generation with Hard Constraints
Summary
The paper presents Neuro-Symbolic Graph Generative Modeling (NSGGM), a framework that enhances molecule generation by integrating symbolic assembly with neural networks, ensuring controllability and compliance with chemical rules.
Why It Matters
This research addresses the limitations of traditional deep learning approaches in molecular and graph generation, providing a method that combines the strengths of neural networks and symbolic reasoning. It offers significant advancements in generating valid and interpretable molecular structures, which is crucial for fields like drug discovery and materials science.
Key Takeaways
- NSGGM combines neural networks with symbolic reasoning for molecule generation.
- The framework ensures compliance with chemical validity and user-defined constraints.
- Introduces a benchmark for evaluating strict rule satisfaction in molecular workflows.
- Demonstrates strong performance in both unconstrained and constrained generation tasks.
- Offers explicit controllability and guarantees that traditional methods lack.
Computer Science > Machine Learning arXiv:2602.16954 (cs) [Submitted on 18 Feb 2026] Title:Neural Proposals, Symbolic Guarantees: Neuro-Symbolic Graph Generation with Hard Constraints Authors:Chuqin Geng, Li Zhang, Mark Zhang, Haolin Ye, Ziyu Zhao, Xujie Si View a PDF of the paper titled Neural Proposals, Symbolic Guarantees: Neuro-Symbolic Graph Generation with Hard Constraints, by Chuqin Geng and 5 other authors View PDF Abstract:We challenge black-box purely deep neural approaches for molecules and graph generation, which are limited in controllability and lack formal guarantees. We introduce Neuro-Symbolic Graph Generative Modeling (NSGGM), a neurosymbolic framework that reapproaches molecule generation as a scaffold and interaction learning task with symbolic assembly. An autoregressive neural model proposes scaffolds and refines interaction signals, and a CPU-efficient SMT solver constructs full graphs while enforcing chemical validity, structural rules, and user-specific constraints, yielding molecules that are correct by construction and interpretable control that pure neural methods cannot provide. NSGGM delivers strong performance on both unconstrained generation and constrained generation tasks, demonstrating that neuro-symbolic modeling can match state-of-the-art generative performance while offering explicit controllability and guarantees. To evaluate more nuanced controllability, we also introduce a Logical-Constraint Molecular Benchmark, designed to test str...
