Controllable Logical Hypothesis Generation for Abductive Reasoning in Knowledge Graphs

2025-05-27

AI Summary

To address the challenges of generating redundant or irrelevant logical hypotheses and lacking controllability in abductive reasoning over knowledge graphs—particularly under single-observation settings—this paper formally introduces the **Controllable Logical Hypothesis Generation** task, enabling users to specify declarative constraints for deriving long-range, complex logical hypotheses. Methodologically, we propose a **Controllable Abductive Generation Paradigm**: (i) a sub-logical decomposition enhancement strategy to mitigate hypothesis space collapse; (ii) smooth semantic rewards (Dice/Overlap) and constraint-following rewards to reduce over-sensitivity to user-specified conditions; and (iii) a two-stage training framework combining supervised learning and reinforcement learning, with constraint-guided decoding. Evaluated on three benchmark datasets, our model achieves significant improvements in constraint adherence and semantic similarity of generated hypotheses, comprehensively outperforming existing baselines.

Technology Category

• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Abductive reasoning in knowledge graphs aims to generate plausible logical hypotheses from observed entities, with broad applications in areas such as clinical diagnosis and scientific discovery. However, due to a lack of controllability, a single observation may yield numerous plausible but redundant or irrelevant hypotheses on large-scale knowledge graphs. To address this limitation, we introduce the task of controllable hypothesis generation to improve the practical utility of abductive reasoning. This task faces two key challenges when controlling for generating long and complex logical hypotheses: hypothesis space collapse and hypothesis oversensitivity. To address these challenges, we propose CtrlHGen, a Controllable logcial Hypothesis Generation framework for abductive reasoning over knowledge graphs, trained in a two-stage paradigm including supervised learning and subsequent reinforcement learning. To mitigate hypothesis space collapse, we design a dataset augmentation strategy based on sub-logical decomposition, enabling the model to learn complex logical structures by leveraging semantic patterns in simpler components. To address hypothesis oversensitivity, we incorporate smoothed semantic rewards including Dice and Overlap scores, and introduce a condition-adherence reward to guide the generation toward user-specified control constraints. Extensive experiments on three benchmark datasets demonstrate that our model not only better adheres to control conditions but also achieves superior semantic similarity performance compared to baselines.

Problem

• Generate controllable logical hypotheses for abductive reasoning
• Address hypothesis space collapse in knowledge graphs
• Mitigate hypothesis oversensitivity with semantic rewards

Innovation

• Two-stage training with supervised and reinforcement learning
• Dataset augmentation via sub-logical decomposition
• Smoothed semantic rewards for condition adherence