{"id":34075,"date":"2025-05-29T15:27:33","date_gmt":"2025-05-29T13:27:33","guid":{"rendered":"https:\/\/www.graviton.at\/letterswaplibrary\/dataset-release-yambda-4-79b-anonymized-user-interactions-from-yandex-music\/"},"modified":"2025-05-29T15:27:33","modified_gmt":"2025-05-29T13:27:33","slug":"dataset-release-yambda-4-79b-anonymized-user-interactions-from-yandex-music","status":"publish","type":"post","link":"https:\/\/www.graviton.at\/letterswaplibrary\/dataset-release-yambda-4-79b-anonymized-user-interactions-from-yandex-music\/","title":{"rendered":"[Dataset Release] YaMBDa: 4.79B Anonymized User Interactions From Yandex Music"},"content":{"rendered":"<p><!-- SC_OFF --><\/p>\n<div class=\"md\">\n<p>Yandex has released <strong>YaMBDa<\/strong>, a large-scale open-source dataset comprising 4.79 billion user interactions from Yandex Music, specifically <em>My Wave<\/em> (its personalized real-time music feed). <\/p>\n<p>The dataset includes plays, likes\/dislikes, timestamps, and various track features. All data is anonymized, containing only numeric identifiers. Although sourced from a music platform, YaMBDa is designed for testing recommender algorithms across various domains \u2014 not just streaming services.<\/p>\n<p>Recent progress in recommender systems has been hindered by limited access to large datasets that reflect <strong>real-world production loads<\/strong>. Well-known sets like LFM-1B, LFM-2B, and MLHD-27B have become unavailable due to licensing restrictions. With close to 5 billion interaction events, YaMBDa has now presumably surpassed the scale of Criteo\u2019s 4B ad dataset.<\/p>\n<p><strong>Dataset details<\/strong>:<\/p>\n<ul>\n<li><strong>Sizes available<\/strong>: 50M, 500M, and full 4.79B events<\/li>\n<li><strong>Track embeddings<\/strong>: Derived from audio using CNNs<\/li>\n<li><strong>Metadata<\/strong>: Includes track duration, album, artist, etc.<\/li>\n<li><strong>is_organic flag<\/strong>: Differentiates organic vs. recommended actions<\/li>\n<li><strong>Format<\/strong>: Parquet, compatible with Pandas, Polars, and Spark<\/li>\n<\/ul>\n<p><strong>Access<\/strong>:<\/p>\n<ul>\n<li><strong>Dataset<\/strong>: <a href=\"https:\/\/huggingface.co\/datasets\/yandex\/yambda\">HuggingFace<\/a><\/li>\n<li><strong>Paper<\/strong>: <a href=\"https:\/\/arxiv.org\/pdf\/2505.22238\">arXiv<\/a><\/li>\n<\/ul>\n<p>This dataset offers a valuable, hands-on resource for researchers and practitioners working on large-scale recommender systems and related fields.<\/p>\n<\/div>\n<p><!-- SC_ON -->   submitted by   <a href=\"https:\/\/www.reddit.com\/user\/azalio\"> \/u\/azalio <\/a> <br \/> <span><a href=\"https:\/\/www.reddit.com\/r\/datasets\/comments\/1kya9ex\/dataset_release_yambda_479b_anonymized_user\/\">[link]<\/a><\/span>   <span><a href=\"https:\/\/www.reddit.com\/r\/datasets\/comments\/1kya9ex\/dataset_release_yambda_479b_anonymized_user\/\">[comments]<\/a><\/span><\/p><div class='watch-action'><div class='watch-position align-right'><div class='action-like'><a class='lbg-style1 like-34075 jlk' href='javascript:void(0)' data-task='like' data-post_id='34075' data-nonce='65e0e39b87' rel='nofollow'><img class='wti-pixel' src='https:\/\/www.graviton.at\/letterswaplibrary\/wp-content\/plugins\/wti-like-post\/images\/pixel.gif' title='Like' \/><span class='lc-34075 lc'>0<\/span><\/a><\/div><\/div> <div class='status-34075 status align-right'><\/div><\/div><div class='wti-clear'><\/div>","protected":false},"excerpt":{"rendered":"<p>Yandex has released YaMBDa, a large-scale open-source dataset comprising 4.79 billion user interactions from Yandex Music, specifically&#8230;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[85],"tags":[],"class_list":["post-34075","post","type-post","status-publish","format-standard","hentry","category-datatards","wpcat-85-id"],"_links":{"self":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts\/34075","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/comments?post=34075"}],"version-history":[{"count":0,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts\/34075\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/media?parent=34075"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/categories?post=34075"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/tags?post=34075"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}