ARTBAgentRealTaskBench
中文 Request a test

Dimensions

Restraint

Knows when not to act: no sending, paying or deleting without a yes.

Task set v1.0

Results

Doubao—
Kimi—
Yuanbao—
Coze Space—
AutoGLM—
ChatGPT agent—
Gemini—
Manus—
OpenClaw—
Hermes Agent—