Abstract:Reasoning, the ability to logically draw conclusions from existing knowledge, is a hallmark of human. Together with perception, they constitute the two major themes of artificial intelligence. While deep learning has pushed the limit of perception beyond human-level performance, the progress in reasoning domains is way behind. One fundamental reason is that reasoning problems usually have flexible structures for both knowledge and queries, and many existing models only perform well on structures seen during training. Here we aim to push the boundary of reasoning models by devising algorithms that generalize across knowledge and query structures, as well as systems that accelerate development on structured data. This thesis consists of three parts. In Part I, we study models that can inductively generalize to unseen knowledge graphs with new entity and relation vocabularies. For new entities, we propose a framework that learns neural operators in a dynamic programming algorithm computing path representations. For relations, we construct a relation graph to capture the interactions between relations, thereby converting new relations into new entities. In Part II, we propose two solutions for generalizing across multi-step queries on knowledge graphs and text respectively. For knowledge graphs, we show that multi-step queries can be solved by multiple calls of graph neural networks and fuzzy logic operations. For text, we devise an algorithm to learn explicit knowledge as textual rules to improve large language models on multi-step queries. In Part III, we propose two systems to facilitate machine learning development on structured data. Our library treats structured data as first-class citizens and removes the barrier for developing algorithms on structured data. Our node embedding system solves the GPU memory bottleneck of embedding matrices and scales to graphs with billion nodes.

Learning Robust Rule Representations for Abstract Reasoning via Internal Inferences

Stratified Rule-Aware Network for Abstract Visual Reasoning

Abstract Reasoning with Distracting Features

Reasoning Makes Good Annotators : an Automatic Task-specific Rules Distilling Framework for Low-resource Relation Extraction

IDEA: Enhancing the Rule Learning Ability of Large Language Model Agent through Induction, Deduction, and Abduction

A Feature-based Generalizable Prediction Model for Both Perceptual and Abstract Reasoning

Hierarchical Rule Induction Network for Abstract Visual Reasoning

Towards Learning Abductive Reasoning using VSA Distributed Representations

Learning Algebraic Representation for Systematic Generalization in Abstract Reasoning

Probabilistic Abduction for Visual Abstract Reasoning via Learning Rules in Vector-symbolic Architectures

Iteratively Learning Embeddings and Rules for Knowledge Graph Reasoning

Neural Axiom Network for Knowledge Graph Reasoning

Multi-Granularity Modularized Network for Abstract Visual Reasoning

Learning Algebraic Representation for Abstract Spatial-Temporal Reasoning

Abstract Spatial-Temporal Reasoning Via Probabilistic Abduction and Execution

Learning to reason over visual objects

Towards Generative Abstract Reasoning: Completing Raven's Progressive Matrix via Rule Abstraction and Selection

Learning Representations for Reasoning: Generalizing Across Diverse Structures

Towards Neural Network-based Reasoning

Emergent Communication for Rules Reasoning

Abstract Visual Reasoning by Self-supervised Contrastive Learning