Hive-ALL: A Semantically Consistent Dataset for Query-Based Universal Sound Separation

Hive-ALL is a dataset for data-efficient query-based universal sound separation, introduced in an arXiv paper from 2026. The dataset is authored by researchers from Tsinghua University, Shanda AI, and Johns Hopkins University. It is hosted on Hugging Face and was last updated on May 26, 2026.

Use Cases

• Training query-based universal sound separation models based on the dataset's semantic consistency.
• Benchmarking data-efficient audio separation algorithms based on the dataset's described properties.
• Researching semantic audio representations for source separation based on the dataset's query-based framework.

Strengths

• Dataset is associated with a 2026 arXiv paper, suggesting recent research relevance.
• Dataset is hosted on Hugging Face and ModelScope, indicating established platform availability.

Limitations

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

Provenance

Source

Tsinghua University, Shanda AI, Johns Hopkins University

Freshness

Last updated 2026-05-26 09:21:07.