Skip to content

Dataset · tencent

Benchmark for AI assistants in group chats

tencent/GroupAssistBench

A benchmark from Tencent and KCL for evaluating AI assistant memory performance in group dialogues.

What's inside

The dataset contains between 1,000 and 10,000 examples in Chinese and English in text and tabular formats. Developed jointly by Tencent Yuanbao and the KCL NLP Group, it tests an assistant's ability to operate in group contexts with multiple memories where simple fact extraction is insufficient. Designed for text generation and question answering tasks.

What you can do with it

  • Benchmarking AI assistants in multi-user chat and long memory scenarios
  • Evaluating RAG systems and memory management mechanisms with multiple participants
  • Testing LLM ability to track context and roles of different conversation participants
Tasks
Question answering, Text generation
Languages
Chinese, English
Size
1K–10K rows
Downloads this month
373
Likes
6

How to load it

from datasets import load_dataset

ds = load_dataset("tencent/GroupAssistBench")