MomentSeeker: A Benchmark for Moment Retrieval Within Long Videos

MomentSeeker is a benchmark dataset for moment retrieval within long videos, released by author avery00 in March 2025. The dataset is hosted on Hugging Face and was last updated on January 14, 2026. It is designed as a comprehensive benchmark and strong baseline for the task.

Use Cases

• Training models for moment retrieval based on textual queries within long videos.
• Benchmarking video-language models on the task of temporal localization.
• Developing strong baseline models for video moment retrieval as described in the associated paper.

Strengths

• Dataset is explicitly designed as a comprehensive benchmark and strong baseline for its task.
• Associated with a research paper, suggesting a structured and validated creation process.
• Last updated on 2026-01-14, indicating recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and sample data are unknown, which may limit suitability assessment.
• License is unknown, requiring verification before use.

Provenance

Source

avery00 on Hugging Face

Collection Method

Annotation data created for the associated research paper.

Time Range

null

Freshness

Last updated 2026-01-14 02:56:51; freshness should be verified.

Geography

null