FAVOR: A Benchmark for Fine-Grained Video Motion Understanding

FAVOR-Bench is a new benchmark for fine-grained video motion understanding released in March 2025. It spans both ego-centric and third-person video perspectives and includes both close-ended QA and open-ended descriptive evaluation tasks. The dataset was created by author zl2048 and last updated on Hugging Face in August 2025.

Use Cases

• Benchmarking MLLM performance on fine-grained motion analysis based on the described evaluation tasks.
• Training models for video question-answering using the close-ended QA tasks mentioned.
• Developing models for open-ended video description generation based on the descriptive tasks included.

Strengths

• Includes both ego-centric and third-person video perspectives, as stated in the description.
• Evaluates models on both close-ended QA and open-ended descriptive tasks, providing a comprehensive assessment.

Limitations

• Description metadata is limited; actual data quality, size, and structure require manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

Hugging Face dataset repository, author zl2048.

Collection Method

Method of data gathering is not specified in the provided input.

Time Range

Benchmark released in March 2025.

Freshness

Last updated 2025-08-01 15:08:12.

Geography

Spatial coverage is not specified in the provided input.