TEA-Dialog

Gold definitionUpdated Apr 2, 2026

Definition

TEA-Dialog is a dataset of tool-enhanced Emotional Support Conversation (ESC) dialogues, designed to facilitate research into building reliable emotional support agents. It provides grounded instrumental support, addressing the limitations of text-only affective support by integrating external tools.

At a glance

Executive summary

TEA-Dialog is a new dataset for training AI systems that give emotional support, but with a twist: these systems can use external tools to provide factual information and avoid making things up. It helps researchers build more reliable and trustworthy AI assistants for sensitive conversations.

TL;DR

TEA-Dialog is a dataset that helps train AI emotional support agents to use tools for factual information, making them more reliable and less prone to errors.

Key points

A dataset of tool-enhanced Emotional Support Conversation (ESC) dialogues.
Solves the problem of factual grounding and hallucination in text-only emotional support systems.
Used by researchers in NLP, dialogue systems, and AI ethics to develop reliable emotional support agents.
Differs from traditional ESC datasets by explicitly incorporating and evaluating external tool use for instrumental support.
Contributes to the growing trend of tool-augmented large language models for sensitive and high-stakes applications.

Use cases

Developing AI chatbots for mental health support that can provide fact-checked advice.
Training virtual assistants to offer reliable guidance in crisis intervention scenarios.
Creating educational support systems that can access and present accurate information during emotional discussions.
Building customer service agents capable of empathetic responses combined with precise, tool-derived solutions.

Also known as

Tool-Enhanced ESC Dialogues Dataset