Dataset · Hugging Face

mlfoundations/dcvlm_pool_medium

Name: dcvlm_pool_medium
Creator: mlfoundations
License: other
Keywords: task_categories:image-text-to-text, task_categories:visual-question-answering, language:en, language:zh, language:de, language:fr, language:ja, language:ko, language:la, language:fi, language:tl, language:bg, language:es, language:nl, license:other

DCVLM-Pool (medium) The raw candidate pool at the medium scale of our DataComp-VLM benchmark: 483,576,747 samples / 41.1 TB across 166 source datasets, as WebDataset tar shards — ≈4× the small pool. This pool is unfiltered and unmixed. It is the input to a data-curation experiment, not a training set. You choose the filters and the mixing ratios, and create another training set. If you instead want a ready-to-train dataset, use dcvlm-baseline-200b (our reference SoTA… See the full description on the dataset page: https://huggingface.co/datasets/mlfoundations/dcvlm_pool_medium.

View on Hugging FaceReport an issue

Author / organization	mlfoundations
License (as declared)	other
Downloads	27.2K (as reported by Hugging Face)
Likes	0
Created	26 July 2026
Last updated at source	17 August 2026

webdatasetmultimodalvision-languagedatacomp

Embed this entry

<iframe src="https://silytics.si/embed/item/2a3bb18d-bda8-4396-9aa7-2fd900bc5d7b" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

More from mlfoundations

mlfoundations/MINT-1T-HTMLDataset
mlfoundations/dclm-baseline-1.0Dataset
mlfoundations/dclm-pool-7b-2xDataset
mlfoundations/datacomp_poolsDataset
mlfoundations/dclm-baseline-1.0-parquetDataset
mlfoundations/dcvlm-baseline-200bDataset

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 14 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.