# EgoExoBench

From the catalogue. Listed by one researcher from one source page.

- id: `egoexobench`
- kind: egocentric
- robot or device: public ego-exo datasets (class: human egocentric)
- how collected: cross-view QA construction
- size as stated: 7,300+ QA pairs; 11 subtasks (per awesome list)
- hours: not stated | hours only claimed: none | episodes: not stated
- year: 2025
- organisation: Yuping He et al.
- licence: not stated (class: not stated)
- access: gated on Hugging Face (class: gated)
- link (paper): https://arxiv.org/abs/2507.18342
- part of: Ego-Exo4D, EgoExoLearn, LEMMA, EgoMe, TF2023 and CVMHAT. Its hours are counted there, not added again to the index's total
- note: Derived benchmark.

Confirm the licence at the link before relying on it.
