Agents & tool use

Mind2Act World

by Yuzhi Huang, Jie Wu, Fanding Huang et al. (Mind2Act World team)

What is Mind2Act World? Mind2Act World is an agents & tool use eval made by Yuzhi Huang, Jie Wu, Fanding Huang et al. (Mind2Act World team). A diagnostic benchmark for robot manipulation in simulation (Isaac Sim) that separates three skills: Mind tasks need decisions based on what the robot saw earlier, Act tasks need precise, well-timed physical execution, and Mind2Act tasks need both in a loop where the outcome of each action shapes the next decision. There are 30 tasks from five application domains, each at three difficulty levels and in clean or randomised scenes, for 180 configurations, scored from physical measurements in the simulator. It evaluates multimodal agents as well as end-to-end vision-language-action and world-action models, and a diagnosis harness attributes failed episodes to Mind, Act or their combination. The authors report that even the strongest agent drops sharply on the combined Mind2Act tasks compared with the single-skill views (arXiv:2611.26444).

Page last updated 2026-10-18 · Evals Review

Editor note

A diagnostic benchmark for robot manipulation in simulation (Isaac Sim) that separates three skills: Mind tasks need decisions based on what the robot saw earlier, Act tasks need precise, well-timed physical execution, and Mind2Act tasks need both in a loop where the outcome of each action shapes the next decision. There are 30 tasks from five application domains, each at three difficulty levels and in clean or randomised scenes, for 180 configurations, scored from physical measurements in the simulator. It evaluates multimodal agents as well as end-to-end vision-language-action and world-action models, and a diagnosis harness attributes failed episodes to Mind, Act or their combination. The authors report that even the strongest agent drops sharply on the combined Mind2Act tasks compared with the single-skill views (arXiv:2611.26444).

What it tests
Embodied memory-based reasoning, physical execution and their closed-loop composition in simulated manipulation.
Original source
mind2act-world.com/
Last updated
arXiv:2611.26444 (19 Nov 2026); repo last pushed 2026-11-20 (GitHub API, read 21 Nov 2026)
License
MIT (code, GitHub API, read 21 Nov 2026)

Facts come only from the linked source (checked 15–16 Oct 2026). Credit belongs to the creators. We don't score or rank evals: how we review.

Reviews

No reviews yet.

Reviews are written by people who sign in with GitHub, one per account per eval. We can't confirm that each reviewer used the eval; we remove reviews that break our rules, never just for being negative, and we never add, buy or edit reviews. Averages, when shown, come only from these reviews.

Sign in with GitHub to review

Related agents & tool use evals

  • GAIA — General assistant skills: browsing, file handling, multi-step reasoning.
  • WebArena — Autonomous web navigation and task completion.
  • OSWorld — Computer-use agents operating real operating systems and apps.
  • Terminal-Bench — Agents working in the shell end to end.

All Agents & tool use evals · Comparison table

Discussion

No one has discussed Mind2Act World yet. Have you run it? Share what it caught, what it missed or how you set it up.

Sign in with GitHub to comment

Be specific and kind. No promotion or affiliate links. Community rules · All discussions