Abstract:Abstract reasoning, as one of the hallmarks of human intelligence, involves collecting information, identifying abstract rules, and applying the rules to solve new problems. Although neural networks have achieved human-level performances in several tasks, the abstract reasoning techniques still far lag behind due to the complexity of learning and applying the logic rules, especially in an unsupervised manner. In this work, we propose a novel framework, ARII, that learns rule representations for Abstract Reasoning via Internal Inferences. The key idea is to repeatedly apply a rule to different instances in hope of having a comprehensive understanding (i.e., representations) of the rule. Specifically, ARII consists of a rule encoder, a reasoner, and an internal referrer. Based on the representations produced by the rule encoder, the reasoner draws the conclusion while the referrer performs internal inferences to regularize rule representations to be robust and generalizable. We evaluate ARII on two benchmark datasets, including PGM and I-RAVEN. We observe that ARII achieves new state-of-the-art records on the majority of the reasoning tasks, including most of the generalization tests in PGM. Our codes are available at https://github.com/Zhangwenbo0324/ARII.

A Kind of Mixed Reasoning in Rule-Based Systems

Hybrid Reasoning Framework Integrating of Rule-Based and Case-Based Reasoning

Local Relations Hold Reasoning

An Implementation of BDI Model Based on Rule Reasoning

The Fuzzy Reasoning Rules Based on Universal Logic.

Research and Design of Reasoning Algorithm with Ontologies and Rules

Argumentation-Based Non-Monotonic Reasoning of Agents

When Do Program-of-Thought Works for Reasoning?

A Hybrid Model for Rule Discovery in Data.

A Probabilistic Soft Logic Reasoning Model with Automatic Rule Learning.

Non-monotonic Reasoning and Argumentation System

An decision-support system based on hybrid reasoning architecture

A Non-revision Reasoning Method for Inconsistent Clausal Beliefs

Automating Defeasible Reasoning in Law

A Model of Reasoning about Knowledge

A Survey of Reasoning with Foundation Models

An Online Incremental Learning Pattern-Based Reasoning System.

An Approach to Robot Commonsense Reasoning

A logical reasoning based decision making method for handling qualitative knowledge

Learning Robust Rule Representations for Abstract Reasoning via Internal Inferences

An improvement to the qualitative interpolative reasoning in sparse rule base