IDAT: A Multi-Modal Dataset and Toolkit for Building and Evaluating Interactive Task-Solving Agents

TL;DR

This paper introduces IDAT, a comprehensive multi-modal dataset and toolkit designed to facilitate the development and evaluation of interactive AI agents capable of understanding and executing grounded natural language instructions in a Minecraft-like environment.

Contribution

The paper presents a scalable data collection tool and an interactive evaluation platform, providing the community with valuable resources for advancing interactive AI agent research.

Findings

Created a dataset with 9,000+ utterances and 1,000+ clarification questions.

Developed a Human-in-the-Loop evaluation platform for multi-turn agent assessment.

Facilitated research in grounded natural language understanding and interactive AI.

Abstract

Seamless interaction between AI agents and humans using natural language remains a key goal in AI research. This paper addresses the challenges of developing interactive agents capable of understanding and executing grounded natural language instructions through the IGLU competition at NeurIPS. Despite advancements, challenges such as a scarcity of appropriate datasets and the need for effective evaluation platforms persist. We introduce a scalable data collection tool for gathering interactive grounded language instructions within a Minecraft-like environment, resulting in a Multi-Modal dataset with around 9,000 utterances and over 1,000 clarification questions. Additionally, we present a Human-in-the-Loop interactive evaluation platform for qualitative analysis and comparison of agent performance through multi-turn communication with human annotators. We offer to the community these assets referred to as IDAT (IGLU Dataset And Toolkit) which aim to advance the development of intelligent, interactive AI agents and provide essential resources for further research.