Large Language Models (LLMs) — News in Games-Based Learning

Can LLMs learn social skills by playing games?

TextArena is an open-source platform developed by A*STAR and IBM to test and improve large language models (LLMs) through over 24 competitive, text-based games. Designed to assess soft skills like negotiation and empathy—areas where LLMs traditionally struggle—the platform promotes real-time evaluation in dynamic environments.

Large Language ModelsDave EngMarch 12, 2026TextArena, Large Language Models (LLMs), Soft skillsComment

How Large Language Models (LLMs) Learn: Playing Games

This article offers a friendly, high-level overview of how Large Language Models (LLMs) like ChatGPT learn. LLMs are statistical models trained on vast amounts of text data, including books, websites, and user content, to understand and generate human-like language.

Players, creators, and AI collaborate to build and expand rich game narratives

In the paper "Player-Driven Emergence in LLM-Driven Game Narrative," the authors explore how large language models (LLMs) like GPT-4 can enhance game narratives by empowering players to contribute creatively. Instead of replacing designers, LLMs enable dynamic interactions with nonplayer characters (NPCs), offering designers valuable feedback.