Summary
Learning logic programs that describe relationships and dependencies is very hard because there are so many possible rules to check. The authors combine large language models (LLMs) with a method that improves rules step-by-step using gradients. Their system, called grasp, lets the language model propose rule candidates guided by gradients, making the search much more efficient. They tested this approach on tasks like predicting molecular toxicity and matching citations, showing it works better than methods that rely only on symbols, neural nets, or language models alone. This approach creates interpretable rule sets that are easier to understand and verify.
What this means in practice
- •For drug discovery teams: Create interpretable predictive models for molecular toxicity by learning weighted relational rules from chemical data.
- •For academic database managers: Improve citation matching accuracy by generating relational rules that capture complex dependencies in bibliography data.
Authors
Saurabh Mathur, Sahil Sidheekh, Bhavan Vasu, Farbod Tavakkoli, Prasad Tadepalli, Kristian Kersting, Sriraam Natarajan
Abstract
Declarative logic programs offer a powerful and interpretable abstraction for encoding relational structure and neurosymbolic reasoning, by expressing dependencies as weighted compositional rules. However, inducing them from data remains fundamentally hard, bottlenecked by the combinatorial explosion of symbolic search spaces. LLMs have recently emerged as powerful hypothesis generators, but when used in isolation, they lack the capacity to do systematic inductive reasoning needed to reliably synthesize valid programs that fit complex relational distributions. We introduce grasp (Gradient-boosted Synthesis of Probabilistic logic programs), a neurosymbolic framework that casts relational structure learning as functional gradient boosting in which the weak learner is a first-order rule and the intractable inner search is delegated to an LLM proposal oracle. We evaluate grasp on four relational benchmarks spanning molecular toxicity prediction (Tox21), mutagenesis, and citation matching (Cora), and show that it improves over purely symbolic, neural, and LLM-based baselines, while producing interpretable weighted rule ensembles. By replacing combinatorial search with gradient-guided LLM hypothesis generation, grasp retains boosting guarantees without sacrificing the transparency of symbolic outputs.