--- title: How a 2021 Quantization Algorithm Quietly Outperforms... description: EDEN, a rotation-based vector quantization algorithm from NeurIPS 2021/ICML 2022, consistently outperforms TurboQuant, a newer method presented at ICLR 2026.... canonical: https://daily.dev/posts/how-a-2021-quantization-algorithm-quietly-outperforms-its-2026-successor-angiwnmdx twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: How a 2021 Quantization Algorithm Quietly Outperforms Its 2026 Successor | daily.dev og:description: EDEN, a rotation-based vector quantization algorithm from NeurIPS 2021/ICML 2022, consistently outperforms TurboQuant, a newer method presented at ICLR 2026.... og:url: https://daily.dev/posts/how-a-2021-quantization-algorithm-quietly-outperforms-its-2026-successor-angiwnmdx og:image: https://api.daily.dev/og/posts/ANgIWNMDx.png og:image:alt: How a 2021 Quantization Algorithm Quietly Outperforms Its 2026 Successor og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # How a 2021 Quantization Algorithm Quietly Outperforms Its 2026 Successor **[Towards Data Science](https://daily.dev/sources/tds)** · 7 min read · 2 upvotes · 0 comments ## Summary EDEN, a rotation-based vector quantization algorithm from NeurIPS 2021/ICML 2022, consistently outperforms TurboQuant, a newer method presented at ICLR 2026. The key difference is a single analytically derived scale parameter: EDEN computes the optimal scale for the known post-rotation Gaussian distribution, while TurboQuant-mse fixes the scale at 1, making it a strictly weaker special case of EDEN. For MSE-targeted compression, EDEN-biased reduces MSE by up to 2.25% at practical bit-widths (2–3 bits). For unbiased estimation (distributed training, approximate attention), EDEN-unbiased outperforms TurboQuant-prod so substantially that 1-bit EDEN-unbiased matches 2-bit TurboQuant-prod accuracy. The comparison holds on TurboQuant's own benchmarks (GloVe, dbpedia embeddings). EDEN implementations are available in PyTorch, TensorFlow, Intel OpenFL, and Google FedJax. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Similar posts on daily.dev - [Qdrant TurboQuant Explained: Is TurboQuant the Silver Bullet?](https://daily.dev/posts/qdrant-turboquant-explained-is-turboquant-the-silver-bullet--nale0jprh) · Towards Data Science · 1 upvotes · 0 comments - [TurboQuant: A First-Principles Walkthrough](https://daily.dev/posts/turboquant-a-first-principles-walkthrough-wemarbhdg) · Hacker News · 0 upvotes · 0 comments - [Effective KV Compression with TurboQuant](https://daily.dev/posts/effective-kv-compression-with-turboquant-nzvsbcxvg) · Machine Learning Mastery · 2 upvotes · 0 comments --- Tags: [#deep-learning](https://daily.dev/tags/deep-learning), [#llm](https://daily.dev/tags/llm), [#machine-learning](https://daily.dev/tags/machine-learning) [View this post on daily.dev](https://daily.dev/posts/how-a-2021-quantization-algorithm-quietly-outperforms-its-2026-successor-angiwnmdx) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"How a 2021 Quantization Algorithm Quietly Outperforms Its 2026 Successor","url":"https://daily.dev/posts/how-a-2021-quantization-algorithm-quietly-outperforms-its-2026-successor-angiwnmdx","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/how-a-2021-quantization-algorithm-quietly-outperforms-its-2026-successor-angiwnmdx"},"datePublished":"2026-05-02T13:40:40.781Z","dateModified":"2026-10-15T02:16:16.303Z","description":"EDEN, a rotation-based vector quantization algorithm from NeurIPS 2021/ICML 2022, consistently outperforms TurboQuant, a newer method presented at ICLR 2026....","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/3b5a503aec77e5f98b665f5ff26f168f?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/3b5a503aec77e5f98b665f5ff26f168f?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Towards Data Science","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Towards Data Science","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/tds","url":"https://daily.dev/sources/tds"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/how-a-2021-quantization-algorithm-quietly-outperforms-its-2026-successor-angiwnmdx","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"deep-learning,llm,machine-learning","timeRequired":"PT7M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Towards Data Science","item":"https://daily.dev/sources/tds"},{"@type":"ListItem","position":3,"name":"How a 2021 Quantization Algorithm Quietly Outperforms Its 2026 Successor"}]} ```