Article

Agents can generalize to novel levels of abstraction with the help of language

2024-10-28

Abstract excerpt

<p>We study abstraction in an emergent communication paradigm. In emergent communication, two artificial neural network agents develop a language while solving a communicative task. In this study, the agents play a concept-level reference game. This means that the speaker agent has to describe a concept to a listener agent, who has to pick the correct target objects that satisfy the concept. Concepts consist of mu...

Topics

Open a Topic to create a Post that cites this publication.

Identifiers and source

Literature Corpus work: 4472ff98-1a58-5ccd-97c3-6447007d3192
DOI: 10.31234/osf.io/km28h

Open publication

Related research

Semantic proximity does not establish scientific evidence.

Click a neighbor to travelStep 1 · 12 closest

Agents can generalize to novel levels of abstraction with the help of languageDOI 10.31234/osf.io/km28hRead

How Can We Characterize Human Generalization and Distinguish it from Generalization in Machines?Semantic proximity Read

The role of pragmatic mechanisms in referential communication and categorization: An emergent communication modelSemantic proximity Read

Assembling hierarchies of action using sequencing and abstraction: studies and models of zero-shot learningSemantic proximity Read

Using Compositionality to Learn Many Categories from Few ExamplesSemantic proximity Read

The Generalization Artist - Or: How Can We Characterize Human Generalization?Semantic proximity Read

Decomposing dynamical subprocesses for compositional generalizationSemantic proximity Read

Toward a reusable architecture for intelligent agentsSemantic proximity Read

Unifying principles of generalization: past, present, and futureSemantic proximity Read

SCALE: Dialogical Cognition as the Algorithmic Engine of Far Transfer (A Target Article for Behavioral and Brain Sciences)Semantic proximity Read