Dataset · internlm
WildClawBench agent benchmark
internlm/WildClawBench
60 practical tasks for evaluating AI agents in a live OpenClaw environment.
What's inside
The dataset contains a benchmark of 60 original tasks for AI agents executed end-to-end in the live OpenClaw open personal assistant environment. Tasks are practical, such as clipping soccer match highlights and scheduling meetings through multi-turn dialogue. Languages include English and Chinese; size is under 1K.
What you can do with it
- Evaluating agentic LLMs on end-to-end tasks
- Testing planning and tool use
- Comparing agents in OpenClaw scenarios
- Tasks
- Visual question answering, Image text → text, Question answering
- Languages
- English, Chinese
- Size
- up to 1K rows
- Downloads this month
- 11.7K
- Likes
- 73
How to load it
from datasets import load_dataset
ds = load_dataset("internlm/WildClawBench")