Update dataset card: queries + ground truth, image join

README.md

@@ -4,53 +4,123 @@ license: cc-by-4.0
 language:
 - en
 ---
-
-# Coyo-VE
+# Coyo-Vector-Embeddings (Coyo-VE)
 
 ## Overview
-Coyo-VE is a large-scale visual-embedding slice of the COYO subset from the [LLaVA-OneVision-1.5-Mid-Training-85M](https://huggingface.co/datasets/mvp-lab/LLaVA-OneVision-1.5-Mid-Training-85M) dataset. Embedded using [Qwen3-VL-Embedding-2B](https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct) (or `qwen3-vl-embedding-2b`), this dataset provides high-dimensional vector representations coupled with multimodal metadata to support web-scale visual and multimodal retrieval benchmarks, vector database indexing, and similarity search.
 
-The dataset includes:
+Coyo-VE is a large-scale visual-text-embedding slice of the COYO subset from the [LLaVA-OneVision-1.5-Mid-Training-85M](https://huggingface.co/datasets/mvp-lab/LLaVA-OneVision-1.5-Mid-Training-85M) dataset, with the original images being sourced from [coyo-700m](https://huggingface.co/datasets/kakaobrain/coyo-700m). Each image-caption pair is jointly embedded using [Qwen3-VL-Embedding-2B](https://huggingface.co/Qwen/Qwen3-VL-Embedding-2B), which accepts mixed image-and-text inputs and encodes their combined visual and textual content into a single **2,048-dimensional dense vector**. This produces a unified multimodal representation for each data point, enabling retrieval based jointly on its visual and textual semantics.
+
+The benchmark includes 25,000 joint image-text queries selected from the ImageNet subset of LLaVA-OneVision-1.5-Mid-Training-85M. These queries model a social-media retrieval scenario in which a user searches for posts related to a given image-text query.
 
-- **Pre-computed multimodal text and visual embeddings** generated with `qwen3-vl-embedding-2b`
-- **Original COYO text payload metadata** retained verbatim with the ability to join images
-- **Scalable parquet structure** optimized for distributed vector database ingestion and similarity benchmarks
+Headline statistics:
+
+| Statistic | Value |
+|---|---|
+| Corpus items (image-caption pairs) | 15,380,795 |
+| Embedding dimensions | 2,048 (dense, `float32`) |
+| Corpus size, with inline images | 5.81 TiB |
+| Queries | 25,000 |
+| Ground truth | exact brute force, k = 1,000, cosine |
 
 ## Dataset Contents
 
+The corpus is derived from the COYO subset of [LLaVA-OneVision-1.5-Mid-Training-85M](https://huggingface.co/datasets/mvp-lab/LLaVA-OneVision-1.5-Mid-Training-85M), which is itself derived from [COYO-700M](https://huggingface.co/datasets/kakaobrain/coyo-700m). Embeddings and ground truth were produced with the `nova-embed` and `nova-bf` components of the [Supernova](https://github.com/qdrant-labs/supernova) pipeline
+
+Each corpus item contains a paired image and textual caption.
+
 Each row contains the document/image payload along with its high-dimensional vector embedding:
 
-| Field | Type | Notes |
-|---|---|---|
-| `id` | `string` | Unique document / image identifier |
-| `text` | `string` | Associated caption or textual content |
-| `qwen_multimodal_embedding` | `float[]` | Multimodal embedding (2048 dim) |
+| **Field**                     | **Type**  | **Notes**                             |
+| ----------------------------- | --------- | ------------------------------------- |
+| `id`                          | `string`  | Unique document / image identifier    |
+| `text`                        | `string`  | Associated caption or textual content |
+| `qwen_multimodal_embedding`   | `float[]` | Multimodal embedding (2048 dim)       |
+
 
 
 ### Embeddings
-Each entry is represented using **[Qwen3-VL-Embedding-2B](https://huggingface.co/Qwen/Qwen3-VL-Embedding-2B)**.
 
-- **Dense:** Multi-modal dense vectors capturing visual and contextual semantic representations.
-- **Metric:** Cosine similarity or dot product on unit-normalized representations.
+Each image-caption pair is jointly embedded using **Qwen3-VL-Embedding-2B**.
+
+- **Modality:** joint image + text (`input_column: text=caption,image=image`, `modality: multimodal`)
+- **Representation:** dense
+- **Dimensions:** 2,048
+- **Datatype:** `float32`
+- **Similarity:** cosine
+
+### Queries and Ground Truth
+
+The benchmark includes 25,000 joint image-text queries and ground truth results at k=1000, together in `queries/queries_gt_k1000.parquet`. Each query consists of an image and associated text that are jointly embedded using the same Qwen3-VL-Embedding-2B representation as the corpus. The queries were selected from the ImageNet subset of LLaVA-OneVision-1.5-Mid-Training-85M, and include the following fields.
+
+
+| Column | Type | Meaning |
+|---|---|---|
+| `query_id` | `string` | query identifier (`q000000` ... `q024999`) |
+| `query` | `string` | query text |
+| `id` | `string` | source image path of the query item |
+| `qwen_multimodal_embedding` | `list<float>` | the 2,048-dim query vector used for the search, unit-norm |
+| `hit_ids` | `list<string>` | 1,000 corpus `id` values, best first |
+| `hit_scores` | `list<float>` | 1,000 cosine scores, descending |
+
+Ground truth is exact: every hit list holds 1,000 unique corpus ids in strictly descending score order, and `hit_ids` joins directly against the corpus `id` column.
+
+
+
+## Downloading and Joining the Images
+
+This release contains embeddings and captions but no image bytes. The images live upstream in [`mvp-lab/LLaVA-OneVision-1.5-Mid-Training-85M`](https://huggingface.co/datasets/mvp-lab/LLaVA-OneVision-1.5-Mid-Training-85M), keyed by the same `id` this release uses, and `join_image_urls.py` fetches them and joins them to the rows for you:
+
+```bash
+# Try it on 1,000 rows first.
+./join_image_urls.py --limit 1000
+
+# The whole corpus (15.4M rows, several TB read from upstream).
+./join_image_urls.py
+
+# The benchmark queries instead of the corpus.
+./join_image_urls.py --what queries --limit 100
+```
+
+The script needs only `pyarrow` and `huggingface_hub`.
+
+It writes the image files into `./images/` and a Parquet — `corpus_with_images.parquet`, or `queries_with_images.parquet` for queries — carrying the release's own columns plus `image_path`:
+
+| Column | Notes |
+|---|---|
+| `id` | this release's row id |
+| `caption` | the recaptioned text (`query_id`, `query` for queries) |
+| `qwen_multimodal_embedding` | the 2,048-dim vector; `--no-embeddings` leaves it out |
+| `image_path` | the image file on disk, or null if the row's image could not be found |
+
+Useful flags: `--images DIR` to write elsewhere, `--limit N` to stop early, `--shards` to point at part of the corpus (say `'hf://datasets/Qdrant/Coyo-VE/data/000/*.parquet'`) or at a local copy.
+
+Only what you ask for is downloaded. Each upstream file's footer statistics rule the file out without reading a single column, so the `id` column is read only for the surviving row groups and the image bytes only where a wanted id actually sits — fetching two images touched columns in 5 of 63 upstream shards.
 
 ### Intended Uses
 
-Coyo-VE is intended for large-scale evaluation of:
+Coyo-VE is intended for evaluating multimodal retrieval systems in workloads where images and text jointly determine semantic similarity.
+
+Potential uses include:
+
+* multimodal vector search
+* image-text retrieval
+* approximate nearest-neighbor search
+* vector database ingestion and indexing
+* distributed retrieval systems
+* retrieval over social-media-style image-text posts
+* large-scale multimodal embedding evaluation
 
-- Visual and multimodal dense vector search
-- Image-text and cross-modal retrieval systems
-- Vector database ingestion, indexing, and ANN benchmark testing
-- Quantization, compression, and memory-footprint evaluation at web scale
-- Distributed retrieval, sharding, and routing strategies
 
 ## Licensing
 
 This dataset slice and its metadata are made available under the **CC-BY-4.0** license, following the upstream license terms of the COYO dataset and LLaVA-OneVision-1.5 release.
 
 ## Acknowledgments
-This dataset would not exist without the upstream work it is built on:
 
-- **MVP Lab** for releasing [LLaVA-OneVision-1.5-Mid-Training-85M](https://huggingface.co/datasets/mvp-lab/LLaVA-OneVision-1.5-Mid-Training-85M)
-- **Qdrant** for vector storage, migration tooling, and host infrastructure
-- **Qwen Team / Alibaba Cloud** for the Qwen-VL embedding models
-- **HuggingFace** for hosting the dataset repository
+We thank [HuggingFace](https://huggingface.co/) for providing a grant to offset a portion of the cost of hosting this dataset. Additionally,  Coyo-VE would not exist without the upstream work it builds on:
+
+* the **MVP-Lab / LLaVA-OneVision-1.5** authors, for assembling and recaptioning the 85M mid-training corpus;
+* **Kakao Brain**, for collecting and releasing COYO-700M;
+* the **Qwen** team, for releasing Qwen3-VL-Embedding-2;
+* the **ImageNet** authors, whose collection supplies the benchmark query items;
+* the **vLLM** project, which serves the embedding model in this pipeline.