EgoLink2026 / README.md
EgoLink's picture
Set dataset split name to Evaluation
ad6059e verified
|
Raw
History Blame Contribute Delete
500 Bytes
---
license: cc-by-nc-sa-4.0
task_categories:
- question-answering
language:
- en
tags:
- egocentric
- social-reasoning
- EgoLink
size_categories:
- 1K<n<10K
configs:
- config_name: default
data_files:
- split: Evaluation
path: Eval.jsonl
---
# EgoLink2026
EgoLink 2026 evaluation benchmark for egocentric social reasoning.
## Dataset Structure
- `Eval.jsonl`: 4030 multiple-choice evaluation questions
- `data/`: 1055 egocentric video files referenced by the `path` field in each sample