Daily-Omni
This repository provides the question-answering metadata of the Daily-Omni benchmark in a format compatible with lmms-eval.
The data is provided as a single parquet file containing only the QA annotations. Since raw videos are not included, please download them from the original release and match them with the QA annotations using video_id.
Task configurations and evaluation scripts are available in the SEATS repository: https://github.com/xxayt/SEATS.
Overview
Daily-Omni is an audio-visual benchmark focused on short, daily-life videos that require joint reasoning over audio and visual modalities. Each sample is a multiple-choice QA over a 30-second clip covering a wide range of everyday topics (lifestyle, cooking, sports, etc.).
| Statistic | Value |
|---|---|
| QA pairs | 1,197 |
| Videos | 30 seconds each |
| Categories | 10 parent / 95 fine-grained |
| Question types | Event Sequence, Context Understanding, Comparative Analysis, Inference, Reasoning, AV Event Alignment, ... |
| Language | English |
| License | CC-BY-NC-SA-4.0 |
License
CC-BY-NC-SA-4.0, inherited from the upstream liarliar/Daily-Omni release. Research / non-commercial use only.