Datasets:

---

PIIR

/

Iceberg-dataset

like 2

Follow

Personalized&Inclusive Intelligence Research 4

This PR adds a comprehensive dataset card for the Iceberg benchmark.

It includes:

• Relevant metadata such as license, task_categories, language, and tags.
• Links to the paper (https://huggingface.co/papers/2512.12980) and the GitHub repository (https://github.com/ZJU-DAILY/Iceberg).
• An introduction to the benchmark, detailing its goals and the Information Loss Funnel.
• Detailed descriptions of the datasets, including an overview table, embedding models, and evaluation tasks.
• A list of supported vector similarity search algorithms.
• A "Quick Start" section with installation instructions and sample usage code to help users get started.
• Relevant images (logo, main diagram, pipeline, results) from the GitHub repository to enhance readability.
• A BibTeX citation for proper academic attribution.