PLM-VideoBench: Human-Annotated Resources for Vision-Language Model Evaluation

PLM-VideoBench is a collection of human-annotated resources for evaluating Vision Language models, focused on detailed video understanding. The dataset includes evaluation data for tasks like FGQA, which probes fine-grained activity understanding through multiple-choice questions. It was authored by Facebook and last updated on May 21, 2025.

Use Cases

• Benchmarking fine-grained video question answering based on the described FGQA task.
• Evaluating vision-language model performance on detailed activity understanding tasks.
• Training or fine-tuning models for multimodal reasoning on video content.

Strengths

• Human-annotated resources ensure a benchmark for detailed video understanding.
• Focus on fine-grained activity understanding provides a specific evaluation challenge.
• Last updated on 2025-05-21, indicating recent maintenance.

Limitations

• Row count, file formats, and column-level documentation are unknown, limiting suitability assessment.
• The dataset description is medium-completeness; actual data quality requires manual inspection after download.

Provenance

Source

facebook

Collection Method

Collection of human-annotated resources, likely for research purposes.

Time Range

null

Freshness

Last updated 2025-05-21 22:55:04.

Geography

null