Emergent language: a survey and taxonomy
🤖 AI Summary
This work addresses the foundational questions in multi-agent reinforcement learning (MARL) concerning language emergence—specifically, its necessary preconditions and measurable criteria. Methodologically, we propose the first unified taxonomy of emergent languages, rigorously distinguishing semantic, syntactic, and functional emergence paradigms; develop a multidimensional analytical framework encompassing agent interaction dynamics, environmental constraints, and representational evolution; and integrate bibliometric analysis, cross-modal empirical experiments, and formal modeling to synthesize over 120 key studies. Our contributions include: (i) a theoretically grounded classification of six canonical emergence mechanisms; (ii) three reproducible, task-agnostic evaluation benchmarks for quantifying linguistic structure and function; and (iii) a comprehensive set of analytical tools and empirically validated standards. Collectively, this work establishes a rigorous foundation—conceptual, methodological, and empirical—for advancing the science of emergent communication in MARL.
Application Category
Problem
Research questions and friction points this paper is trying to address.
Explores language emergence in multi-agent reinforcement learning systems.
Reviews prerequisites and success criteria for emergent language in AI.
Identifies research gaps and evaluates methods in emergent language studies.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Multi-agent reinforcement learning for language emergence
Extending beyond statistical NLP representations
Comprehensive review of 181 emergent language studies
🔎 Similar Papers
No similar papers found.
