On Semantic Cognition, Inductive Generalization, and Language Models

TL;DR

This paper explores how language models develop semantic understanding and generalization abilities by applying cognitive science insights into inductive reasoning, aiming to bridge human-like concept learning with neural network behavior.

Contribution

It introduces a framework for analyzing semantic inductive generalization in language models inspired by human inductive reasoning phenomena.

Findings

Language models exhibit inductive reasoning behaviors similar to humans.

Semantic generalization in LMs can be analyzed through cognitive science phenomena.

Inductive dynamics relate to the structure of learned conceptual representations.

Abstract

My doctoral research focuses on understanding semantic knowledge in neural network models trained solely to predict natural language (referred to as language models, or LMs), by drawing on insights from the study of concepts and categories grounded in cognitive science. I propose a framework inspired by 'inductive reasoning,' a phenomenon that sheds light on how humans utilize background knowledge to make inductive leaps and generalize from new pieces of information about concepts and their properties. Drawing from experiments that study inductive reasoning, I propose to analyze semantic inductive generalization in LMs using phenomena observed in human-induction literature, investigate inductive behavior on tasks such as implicit reasoning and emergent feature recognition, and analyze and relate induction dynamics to the learned conceptual representation space.