LongAV-Compass: Evaluation Data for Multimodal Audio-Video Generation

LongAV-Compass provides prompt and event-level evaluation data for three multimodal generation tasks. The dataset covers text-to-audio-video, image-to-audio-video, and video-to-audio-video generation. It was authored by TengfeiLiuCoder and last updated on Hugging Face in May 2026.

Use Cases

• Benchmarking text-to-audio-video models based on prompt-level evaluation data.
• Evaluating image-to-audio-video generation systems using the provided I2AV reference files.
• Assessing video-to-audio-video generation performance on event-level tasks.
• Training multimodal evaluation metrics using the structured prompt-response pairs.

Strengths

• Covers three distinct multimodal generation tasks: T2AV, I2AV, and V2AV.
• Includes both JSON files for structured data and a Hugging Face Viewer friendly sample table.
• Last updated on 2026-05-25, indicating recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface

Freshness

Last updated 2026-05-25 10:09:30.