MM-ToolSandBox Benchmark

What is MM-ToolSandBox?

MM-ToolSandBox is a stateful sandbox benchmark for visually grounded tool-calling agents. It has 258 human-verified scenarios across 500+ tools in 16 domains…

Released

2026-07-13

Evaluates

General AI, Vision & 3D, cs.CV

Openness

unknown

Importer

Claire Radar

Review status

ai-reviewed

Reported scores

0

Source provenance

• Original evidence https://arxiv.org/abs/2607.11818

MM-ToolSandBox paper and code

• Paper https://arxiv.org/abs/2607.11818
• Code https://github.com/apple/ml-mmtoolsandbox

Which sources cite MM-ToolSandBox?

• MM-ToolSandBox

Related General AI, Vision & 3D, cs.CV benchmarks

• $\tau^\tau$-Bench: An Environment for End-To-End, Realistic Agent Construction
• 360CityArena
• 3DCodeBench
• 5-Dialects-BN
• A3S-Bench
• AARRI-Bench
• ActBench
• ActiveSciBench